linux/fs/coredump.c

1258 lines
31 KiB
C
Raw Normal View History

License cleanup: add SPDX GPL-2.0 license identifier to files with no license Many source files in the tree are missing licensing information, which makes it harder for compliance tools to determine the correct license. By default all files without license information are under the default license of the kernel, which is GPL version 2. Update the files which contain no license information with the 'GPL-2.0' SPDX license identifier. The SPDX identifier is a legally binding shorthand, which can be used instead of the full boiler plate text. This patch is based on work done by Thomas Gleixner and Kate Stewart and Philippe Ombredanne. How this work was done: Patches were generated and checked against linux-4.14-rc6 for a subset of the use cases: - file had no licensing information it it. - file was a */uapi/* one with no licensing information in it, - file was a */uapi/* one with existing licensing information, Further patches will be generated in subsequent months to fix up cases where non-standard license headers were used, and references to license had to be inferred by heuristics based on keywords. The analysis to determine which SPDX License Identifier to be applied to a file was done in a spreadsheet of side by side results from of the output of two independent scanners (ScanCode & Windriver) producing SPDX tag:value files created by Philippe Ombredanne. Philippe prepared the base worksheet, and did an initial spot review of a few 1000 files. The 4.13 kernel was the starting point of the analysis with 60,537 files assessed. Kate Stewart did a file by file comparison of the scanner results in the spreadsheet to determine which SPDX license identifier(s) to be applied to the file. She confirmed any determination that was not immediately clear with lawyers working with the Linux Foundation. Criteria used to select files for SPDX license identifier tagging was: - Files considered eligible had to be source code files. - Make and config files were included as candidates if they contained >5 lines of source - File already had some variant of a license header in it (even if <5 lines). All documentation files were explicitly excluded. The following heuristics were used to determine which SPDX license identifiers to apply. - when both scanners couldn't find any license traces, file was considered to have no license information in it, and the top level COPYING file license applied. For non */uapi/* files that summary was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 11139 and resulted in the first patch in this series. If that file was a */uapi/* path one, it was "GPL-2.0 WITH Linux-syscall-note" otherwise it was "GPL-2.0". Results of that was: SPDX license identifier # files ---------------------------------------------------|------- GPL-2.0 WITH Linux-syscall-note 930 and resulted in the second patch in this series. - if a file had some form of licensing information in it, and was one of the */uapi/* ones, it was denoted with the Linux-syscall-note if any GPL family license was found in the file or had no licensing in it (per prior point). Results summary: SPDX license identifier # files ---------------------------------------------------|------ GPL-2.0 WITH Linux-syscall-note 270 GPL-2.0+ WITH Linux-syscall-note 169 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) 21 ((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause) 17 LGPL-2.1+ WITH Linux-syscall-note 15 GPL-1.0+ WITH Linux-syscall-note 14 ((GPL-2.0+ WITH Linux-syscall-note) OR BSD-3-Clause) 5 LGPL-2.0+ WITH Linux-syscall-note 4 LGPL-2.1 WITH Linux-syscall-note 3 ((GPL-2.0 WITH Linux-syscall-note) OR MIT) 3 ((GPL-2.0 WITH Linux-syscall-note) AND MIT) 1 and that resulted in the third patch in this series. - when the two scanners agreed on the detected license(s), that became the concluded license(s). - when there was disagreement between the two scanners (one detected a license but the other didn't, or they both detected different licenses) a manual inspection of the file occurred. - In most cases a manual inspection of the information in the file resulted in a clear resolution of the license that should apply (and which scanner probably needed to revisit its heuristics). - When it was not immediately clear, the license identifier was confirmed with lawyers working with the Linux Foundation. - If there was any question as to the appropriate license identifier, the file was flagged for further research and to be revisited later in time. In total, over 70 hours of logged manual review was done on the spreadsheet to determine the SPDX license identifiers to apply to the source files by Kate, Philippe, Thomas and, in some cases, confirmation by lawyers working with the Linux Foundation. Kate also obtained a third independent scan of the 4.13 code base from FOSSology, and compared selected files where the other two scanners disagreed against that SPDX file, to see if there was new insights. The Windriver scanner is based on an older version of FOSSology in part, so they are related. Thomas did random spot checks in about 500 files from the spreadsheets for the uapi headers and agreed with SPDX license identifier in the files he inspected. For the non-uapi files Thomas did random spot checks in about 15000 files. In initial set of patches against 4.14-rc6, 3 files were found to have copy/paste license identifier errors, and have been fixed to reflect the correct identifier. Additionally Philippe spent 10 hours this week doing a detailed manual inspection and review of the 12,461 patched files from the initial patch version early this week with: - a full scancode scan run, collecting the matched texts, detected license ids and scores - reviewing anything where there was a license detected (about 500+ files) to ensure that the applied SPDX license was correct - reviewing anything where there was no detection but the patch license was not GPL-2.0 WITH Linux-syscall-note to ensure that the applied SPDX license was correct This produced a worksheet with 20 files needing minor correction. This worksheet was then exported into 3 different .csv files for the different types of files to be modified. These .csv files were then reviewed by Greg. Thomas wrote a script to parse the csv files and add the proper SPDX tag to the file, in the format that the file expected. This script was further refined by Greg based on the output to detect more types of files automatically and to distinguish between header and source .c files (which need different comment types.) Finally Greg ran the script using the .csv files to generate the patches. Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Reviewed-by: Philippe Ombredanne <pombredanne@nexb.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2017-11-01 14:07:57 +00:00
// SPDX-License-Identifier: GPL-2.0
#include <linux/slab.h>
#include <linux/file.h>
#include <linux/fdtable.h>
#include <linux/freezer.h>
#include <linux/mm.h>
#include <linux/stat.h>
#include <linux/fcntl.h>
#include <linux/swap.h>
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
#include <linux/ctype.h>
#include <linux/string.h>
#include <linux/init.h>
#include <linux/pagemap.h>
#include <linux/perf_event.h>
#include <linux/highmem.h>
#include <linux/spinlock.h>
#include <linux/key.h>
#include <linux/personality.h>
#include <linux/binfmts.h>
#include <linux/coredump.h>
#include <linux/sched/coredump.h>
#include <linux/sched/signal.h>
#include <linux/sched/task_stack.h>
#include <linux/utsname.h>
#include <linux/pid_namespace.h>
#include <linux/module.h>
#include <linux/namei.h>
#include <linux/mount.h>
#include <linux/security.h>
#include <linux/syscalls.h>
#include <linux/tsacct_kern.h>
#include <linux/cn_proc.h>
#include <linux/audit.h>
#include <linux/kmod.h>
#include <linux/fsnotify.h>
#include <linux/fs_struct.h>
#include <linux/pipe_fs_i.h>
#include <linux/oom.h>
#include <linux/compat.h>
2016-03-22 21:25:36 +00:00
#include <linux/fs.h>
#include <linux/path.h>
#include <linux/timekeeping.h>
#include <linux/sysctl.h>
#include <linux/elf.h>
#include <linux/uaccess.h>
#include <asm/mmu_context.h>
#include <asm/tlb.h>
#include <asm/exec.h>
#include <trace/events/task.h>
#include "internal.h"
#include <trace/events/sched.h>
static bool dump_vma_snapshot(struct coredump_params *cprm);
static void free_vma_snapshot(struct coredump_params *cprm);
#define CORE_FILE_NOTE_SIZE_DEFAULT (4*1024*1024)
/* Define a reasonable max cap */
#define CORE_FILE_NOTE_SIZE_MAX (16*1024*1024)
static int core_uses_pid;
static unsigned int core_pipe_limit;
static char core_pattern[CORENAME_MAX_SIZE] = "core";
static int core_name_size = CORENAME_MAX_SIZE;
unsigned int core_file_note_size_limit = CORE_FILE_NOTE_SIZE_DEFAULT;
struct core_name {
char *corename;
int used, size;
};
static int expand_corename(struct core_name *cn, int size)
{
char *corename;
size = kmalloc_size_roundup(size);
corename = krealloc(cn->corename, size, GFP_KERNEL);
if (!corename)
return -ENOMEM;
if (size > core_name_size) /* racy but harmless */
core_name_size = size;
cn->size = size;
cn->corename = corename;
return 0;
}
static __printf(2, 0) int cn_vprintf(struct core_name *cn, const char *fmt,
va_list arg)
{
int free, need;
coredump: fix va_list corruption A va_list needs to be copied in case it needs to be used twice. Thanks to Hugh for debugging this issue, leading to various panics. Tested: lpq84:~# echo "|/foobar12345 %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h" >/proc/sys/kernel/core_pattern 'produce_core' is simply : main() { *(int *)0 = 1;} lpq84:~# ./produce_core Segmentation fault (core dumped) lpq84:~# dmesg | tail -1 [ 614.352947] Core dump to |/foobar12345 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 (null) pipe failed Notice the last argument was replaced by a NULL (we were lucky enough to not crash, but do not try this on your production machine !) After fix : lpq83:~# echo "|/foobar12345 %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h" >/proc/sys/kernel/core_pattern lpq83:~# ./produce_core Segmentation fault lpq83:~# dmesg | tail -1 [ 740.800441] Core dump to |/foobar12345 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 pipe failed Fixes: 5fe9d8ca21cc ("coredump: cn_vprintf() has no reason to call vsnprintf() twice") Signed-off-by: Eric Dumazet <edumazet@google.com> Diagnosed-by: Hugh Dickins <hughd@google.com> Acked-by: Oleg Nesterov <oleg@redhat.com> Cc: Neil Horman <nhorman@tuxdriver.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: stable@vger.kernel.org # 3.11+ Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-19 17:15:07 +00:00
va_list arg_copy;
again:
free = cn->size - cn->used;
coredump: fix va_list corruption A va_list needs to be copied in case it needs to be used twice. Thanks to Hugh for debugging this issue, leading to various panics. Tested: lpq84:~# echo "|/foobar12345 %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h" >/proc/sys/kernel/core_pattern 'produce_core' is simply : main() { *(int *)0 = 1;} lpq84:~# ./produce_core Segmentation fault (core dumped) lpq84:~# dmesg | tail -1 [ 614.352947] Core dump to |/foobar12345 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 lpq84 (null) pipe failed Notice the last argument was replaced by a NULL (we were lucky enough to not crash, but do not try this on your production machine !) After fix : lpq83:~# echo "|/foobar12345 %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h %h" >/proc/sys/kernel/core_pattern lpq83:~# ./produce_core Segmentation fault lpq83:~# dmesg | tail -1 [ 740.800441] Core dump to |/foobar12345 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 lpq83 pipe failed Fixes: 5fe9d8ca21cc ("coredump: cn_vprintf() has no reason to call vsnprintf() twice") Signed-off-by: Eric Dumazet <edumazet@google.com> Diagnosed-by: Hugh Dickins <hughd@google.com> Acked-by: Oleg Nesterov <oleg@redhat.com> Cc: Neil Horman <nhorman@tuxdriver.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: stable@vger.kernel.org # 3.11+ Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-04-19 17:15:07 +00:00
va_copy(arg_copy, arg);
need = vsnprintf(cn->corename + cn->used, free, fmt, arg_copy);
va_end(arg_copy);
if (need < free) {
cn->used += need;
return 0;
}
if (!expand_corename(cn, cn->size + need - free + 1))
goto again;
return -ENOMEM;
}
static __printf(2, 3) int cn_printf(struct core_name *cn, const char *fmt, ...)
{
va_list arg;
int ret;
va_start(arg, fmt);
ret = cn_vprintf(cn, fmt, arg);
va_end(arg);
return ret;
}
static __printf(2, 3)
int cn_esc_printf(struct core_name *cn, const char *fmt, ...)
{
int cur = cn->used;
va_list arg;
int ret;
va_start(arg, fmt);
ret = cn_vprintf(cn, fmt, arg);
va_end(arg);
if (ret == 0) {
/*
* Ensure that this coredump name component can't cause the
* resulting corefile path to consist of a ".." or ".".
*/
if ((cn->used - cur == 1 && cn->corename[cur] == '.') ||
(cn->used - cur == 2 && cn->corename[cur] == '.'
&& cn->corename[cur+1] == '.'))
cn->corename[cur] = '!';
/*
* Empty names are fishy and could be used to create a "//" in a
* corefile name, causing the coredump to happen one directory
* level too high. Enforce that all components of the core
* pattern are at least one character long.
*/
if (cn->used == cur)
ret = cn_printf(cn, "!");
}
for (; cur < cn->used; ++cur) {
if (cn->corename[cur] == '/')
cn->corename[cur] = '!';
}
return ret;
}
static int cn_print_exe_file(struct core_name *cn, bool name_only)
{
struct file *exe_file;
char *pathbuf, *path, *ptr;
int ret;
exe_file = get_mm_exe_file(current->mm);
if (!exe_file)
return cn_esc_printf(cn, "%s (path unknown)", current->comm);
mm: treewide: remove GFP_TEMPORARY allocation flag GFP_TEMPORARY was introduced by commit e12ba74d8ff3 ("Group short-lived and reclaimable kernel allocations") along with __GFP_RECLAIMABLE. It's primary motivation was to allow users to tell that an allocation is short lived and so the allocator can try to place such allocations close together and prevent long term fragmentation. As much as this sounds like a reasonable semantic it becomes much less clear when to use the highlevel GFP_TEMPORARY allocation flag. How long is temporary? Can the context holding that memory sleep? Can it take locks? It seems there is no good answer for those questions. The current implementation of GFP_TEMPORARY is basically GFP_KERNEL | __GFP_RECLAIMABLE which in itself is tricky because basically none of the existing caller provide a way to reclaim the allocated memory. So this is rather misleading and hard to evaluate for any benefits. I have checked some random users and none of them has added the flag with a specific justification. I suspect most of them just copied from other existing users and others just thought it might be a good idea to use without any measuring. This suggests that GFP_TEMPORARY just motivates for cargo cult usage without any reasoning. I believe that our gfp flags are quite complex already and especially those with highlevel semantic should be clearly defined to prevent from confusion and abuse. Therefore I propose dropping GFP_TEMPORARY and replace all existing users to simply use GFP_KERNEL. Please note that SLAB users with shrinkers will still get __GFP_RECLAIMABLE heuristic and so they will be placed properly for memory fragmentation prevention. I can see reasons we might want some gfp flag to reflect shorterm allocations but I propose starting from a clear semantic definition and only then add users with proper justification. This was been brought up before LSF this year by Matthew [1] and it turned out that GFP_TEMPORARY really doesn't have a clear semantic. It seems to be a heuristic without any measured advantage for most (if not all) its current users. The follow up discussion has revealed that opinions on what might be temporary allocation differ a lot between developers. So rather than trying to tweak existing users into a semantic which they haven't expected I propose to simply remove the flag and start from scratch if we really need a semantic for short term allocations. [1] http://lkml.kernel.org/r/20170118054945.GD18349@bombadil.infradead.org [akpm@linux-foundation.org: fix typo] [akpm@linux-foundation.org: coding-style fixes] [sfr@canb.auug.org.au: drm/i915: fix up] Link: http://lkml.kernel.org/r/20170816144703.378d4f4d@canb.auug.org.au Link: http://lkml.kernel.org/r/20170728091904.14627-1-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au> Acked-by: Mel Gorman <mgorman@suse.de> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Matthew Wilcox <willy@infradead.org> Cc: Neil Brown <neilb@suse.de> Cc: "Theodore Ts'o" <tytso@mit.edu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-13 23:28:29 +00:00
pathbuf = kmalloc(PATH_MAX, GFP_KERNEL);
if (!pathbuf) {
ret = -ENOMEM;
goto put_exe_file;
}
path = file_path(exe_file, pathbuf, PATH_MAX);
if (IS_ERR(path)) {
ret = PTR_ERR(path);
goto free_buf;
}
if (name_only) {
ptr = strrchr(path, '/');
if (ptr)
path = ptr + 1;
}
ret = cn_esc_printf(cn, "%s", path);
free_buf:
kfree(pathbuf);
put_exe_file:
fput(exe_file);
return ret;
}
/* format_corename will inspect the pattern parameter, and output a
* name into corename, which must have space for at least
* CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
*/
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
static int format_corename(struct core_name *cn, struct coredump_params *cprm,
size_t **argv, int *argc)
{
const struct cred *cred = current_cred();
const char *pat_ptr = core_pattern;
int ispipe = (*pat_ptr == '|');
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
bool was_space = false;
int pid_in_pattern = 0;
int err = 0;
cn->used = 0;
cn->corename = NULL;
if (expand_corename(cn, core_name_size))
return -ENOMEM;
cn->corename[0] = '\0';
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
if (ispipe) {
int argvs = sizeof(core_pattern) / 2;
(*argv) = kmalloc_array(argvs, sizeof(**argv), GFP_KERNEL);
if (!(*argv))
return -ENOMEM;
(*argv)[(*argc)++] = 0;
++pat_ptr;
if (!(*pat_ptr))
return -ENOMEM;
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
}
/* Repeat as long as we have more pattern to process and more output
space */
while (*pat_ptr) {
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
/*
* Split on spaces before doing template expansion so that
* %e and %E don't get split if they have spaces in them
*/
if (ispipe) {
if (isspace(*pat_ptr)) {
if (cn->used != 0)
was_space = true;
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
pat_ptr++;
continue;
} else if (was_space) {
was_space = false;
err = cn_printf(cn, "%c", '\0');
if (err)
return err;
(*argv)[(*argc)++] = cn->used;
}
}
if (*pat_ptr != '%') {
err = cn_printf(cn, "%c", *pat_ptr++);
} else {
switch (*++pat_ptr) {
/* single % at the end, drop that */
case 0:
goto out;
/* Double percent, output one percent */
case '%':
err = cn_printf(cn, "%c", '%');
break;
/* pid */
case 'p':
pid_in_pattern = 1;
err = cn_printf(cn, "%d",
task_tgid_vnr(current));
break;
/* global pid */
case 'P':
err = cn_printf(cn, "%d",
task_tgid_nr(current));
break;
coredump: add %i/%I in core_pattern to report the tid of the crashed thread format_corename() can only pass the leader's pid to the core handler, but there is no simple way to figure out which thread originated the coredump. As Jan explains, this also means that there is no simple way to create the backtrace of the crashed process: As programs are mostly compiled with implicit gcc -fomit-frame-pointer one needs program's .eh_frame section (equivalently PT_GNU_EH_FRAME segment) or .debug_frame section. .debug_frame usually is present only in separate debug info files usually not even installed on the system. While .eh_frame is a part of the executable/library (and it is even always mapped for C++ exceptions unwinding) it no longer has to be present anywhere on the disk as the program could be upgraded in the meantime and the running instance has its executable file already unlinked from disk. One possibility is to echo 0x3f >/proc/*/coredump_filter and dump all the file-backed memory including the executable's .eh_frame section. But that can create huge core files, for example even due to mmapped data files. Other possibility would be to read .eh_frame from /proc/PID/mem at the core_pattern handler time of the core dump. For the backtrace one needs to read the register state first which can be done from core_pattern handler: ptrace(PTRACE_SEIZE, tid, 0, PTRACE_O_TRACEEXIT) close(0); // close pipe fd to resume the sleeping dumper waitpid(); // should report EXIT PTRACE_GETREGS or other requests The remaining problem is how to get the 'tid' value of the crashed thread. It could be read from the first NT_PRSTATUS note of the core file but that makes the core_pattern handler complicated. Unfortunately %t is already used so this patch uses %i/%I. Automatic Bug Reporting Tool (https://github.com/abrt/abrt/wiki/overview) is experimenting with this. It is using the elfutils (https://fedorahosted.org/elfutils/) unwinder for generating the backtraces. Apart from not needing matching executables as mentioned above, another advantage is that we can get the backtrace without saving the core (which might be quite large) to disk. [mmilata@redhat.com: final paragraph of changelog] Signed-off-by: Jan Kratochvil <jan.kratochvil@redhat.com> Signed-off-by: Oleg Nesterov <oleg@redhat.com> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: Jan Kratochvil <jan.kratochvil@redhat.com> Cc: Mark Wielaard <mjw@redhat.com> Cc: Martin Milata <mmilata@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-13 22:53:35 +00:00
case 'i':
err = cn_printf(cn, "%d",
task_pid_vnr(current));
break;
case 'I':
err = cn_printf(cn, "%d",
task_pid_nr(current));
break;
/* uid */
case 'u':
err = cn_printf(cn, "%u",
from_kuid(&init_user_ns,
cred->uid));
break;
/* gid */
case 'g':
err = cn_printf(cn, "%u",
from_kgid(&init_user_ns,
cred->gid));
break;
case 'd':
err = cn_printf(cn, "%d",
__get_dumpable(cprm->mm_flags));
break;
/* signal that caused the coredump */
case 's':
err = cn_printf(cn, "%d",
cprm->siginfo->si_signo);
break;
/* UNIX time of coredump */
case 't': {
time64_t time;
time = ktime_get_real_seconds();
err = cn_printf(cn, "%lld", time);
break;
}
/* hostname */
case 'h':
down_read(&uts_sem);
err = cn_esc_printf(cn, "%s",
utsname()->nodename);
up_read(&uts_sem);
break;
/* executable, could be changed by prctl PR_SET_NAME etc */
case 'e':
err = cn_esc_printf(cn, "%s", current->comm);
break;
/* file name of executable */
case 'f':
err = cn_print_exe_file(cn, true);
break;
case 'E':
err = cn_print_exe_file(cn, false);
break;
/* core limit size */
case 'c':
err = cn_printf(cn, "%lu",
rlimit(RLIMIT_CORE));
break;
/* CPU the task ran on */
case 'C':
err = cn_printf(cn, "%d", cprm->cpu);
break;
default:
break;
}
++pat_ptr;
}
if (err)
return err;
}
out:
/* Backward compatibility with core_uses_pid:
*
* If core_pattern does not include a %p (as is the default)
* and core_uses_pid is set, then .%pid will be appended to
* the filename. Do not do this for piped commands. */
if (!ispipe && !pid_in_pattern && core_uses_pid) {
err = cn_printf(cn, ".%d", task_tgid_vnr(current));
if (err)
return err;
}
return ispipe;
}
static int zap_process(struct signal_struct *signal, int exit_code)
{
struct task_struct *t;
int nr = 0;
signal->flags = SIGNAL_GROUP_EXIT;
signal->group_exit_code = exit_code;
signal->group_stop_count = 0;
__for_each_thread(signal, t) {
task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
coredump: Don't perform any cleanups before dumping core Rename coredump_exit_mm to coredump_task_exit and call it from do_exit before PTRACE_EVENT_EXIT, and before any cleanup work for a task happens. This ensures that an accurate copy of the process can be captured in the coredump as no cleanup for the process happens before the coredump completes. This also ensures that PTRACE_EVENT_EXIT will not be visited by any thread until the coredump is complete. Add a new flag PF_POSTCOREDUMP so that tasks that have passed through coredump_task_exit can be recognized and ignored in zap_process. Now that all of the coredumping happens before exit_mm remove code to test for a coredump in progress from mm_release. Replace "may_ptrace_stop()" with a simple test of "current->ptrace". The other tests in may_ptrace_stop all concern avoiding stopping during a coredump. These tests are no longer necessary as it is now guaranteed that fatal_signal_pending will be set if the code enters ptrace_stop during a coredump. The code in ptrace_stop is guaranteed not to stop if fatal_signal_pending returns true. Until this change "ptrace_event(PTRACE_EVENT_EXIT)" could call ptrace_stop without fatal_signal_pending being true, as signals are dequeued in get_signal before calling do_exit. This is no longer an issue as "ptrace_event(PTRACE_EVENT_EXIT)" is no longer reached until after the coredump completes. Link: https://lkml.kernel.org/r/874kaax26c.fsf@disp2133 Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2021-09-01 16:33:50 +00:00
if (t != current && !(t->flags & PF_POSTCOREDUMP)) {
sigaddset(&t->pending.signal, SIGKILL);
signal_wake_up(t, 1);
nr++;
}
}
return nr;
}
coredump: Limit coredumps to a single thread group Today when a signal is delivered with a handler of SIG_DFL whose default behavior is to generate a core dump not only that process but every process that shares the mm is killed. In the case of vfork this looks like a real world problem. Consider the following well defined sequence. if (vfork() == 0) { execve(...); _exit(EXIT_FAILURE); } If a signal that generates a core dump is received after vfork but before the execve changes the mm the process that called vfork will also be killed (as the mm is shared). Similarly if the execve fails after the point of no return the kernel delivers SIGSEGV which will kill both the exec'ing process and because the mm is shared the process that called vfork as well. As far as I can tell this behavior is a violation of people's reasonable expectations, POSIX, and is unnecessarily fragile when the system is low on memory. Solve this by making a userspace visible change to only kill a single process/thread group. This is possible because Jann Horn recently modified[1] the coredump code so that the mm can safely be modified while the coredump is happening. With LinuxThreads long gone I don't expect anyone to have a notice this behavior change in practice. To accomplish this move the core_state pointer from mm_struct to signal_struct, which allows different thread groups to coredump simultatenously. In zap_threads remove the work to kill anything except for the current thread group. v2: Remove core_state from the VM_BUG_ON_MM print to fix compile failure when CONFIG_DEBUG_VM is enabled. Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> [1] a07279c9a8cd ("binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot") Fixes: d89f3847def4 ("[PATCH] thread-aware coredumps, 2.5.43-C3") History-tree: git://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git Link: https://lkml.kernel.org/r/87y27mvnke.fsf@disp2133 Link: https://lkml.kernel.org/r/20211007144701.67592574@canb.auug.org.au Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2021-09-22 16:24:02 +00:00
static int zap_threads(struct task_struct *tsk,
coredump: only SIGKILL should interrupt the coredumping task There are 2 well known and ancient problems with coredump/signals, and a lot of related bug reports: - do_coredump() clears TIF_SIGPENDING but of course this can't help if, say, SIGCHLD comes after that. In this case the coredump can fail unexpectedly. See for example wait_for_dump_helper()->signal_pending() check but there are other reasons. - At the same time, dumping a huge core on the slow media can take a lot of time/resources and there is no way to kill the coredumping task reliably. In particular this is not oom_kill-friendly. This patch tries to fix the 1st problem, and makes the preparation for the next changes. We add the new SIGNAL_GROUP_COREDUMP flag set by zap_threads() to indicate that this process dumps the core. prepare_signal() checks this flag and nacks any signal except SIGKILL. Note that this check tries to be conservative, in the long term we should probably treat the SIGNAL_GROUP_EXIT case equally but this needs more discussion. See marc.info/?l=linux-kernel&m=120508897917439 Notes: - recalc_sigpending() doesn't check SIGNAL_GROUP_COREDUMP. The patch assumes that dump_write/etc paths should never call it, but we can change it as well. - There is another source of TIF_SIGPENDING, freezer. This will be addressed separately. Signed-off-by: Oleg Nesterov <oleg@redhat.com> Tested-by: Mandeep Singh Baines <msb@chromium.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Neil Horman <nhorman@redhat.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Roland McGrath <roland@hack.frob.com> Cc: Tejun Heo <tj@kernel.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 22:28:10 +00:00
struct core_state *core_state, int exit_code)
{
struct signal_struct *signal = tsk->signal;
int nr = -EAGAIN;
spin_lock_irq(&tsk->sighand->siglock);
if (!(signal->flags & SIGNAL_GROUP_EXIT) && !signal->group_exec_task) {
/* Allow SIGKILL, see prepare_signal() */
signal->core_state = core_state;
nr = zap_process(signal, exit_code);
coredump: only SIGKILL should interrupt the coredumping task There are 2 well known and ancient problems with coredump/signals, and a lot of related bug reports: - do_coredump() clears TIF_SIGPENDING but of course this can't help if, say, SIGCHLD comes after that. In this case the coredump can fail unexpectedly. See for example wait_for_dump_helper()->signal_pending() check but there are other reasons. - At the same time, dumping a huge core on the slow media can take a lot of time/resources and there is no way to kill the coredumping task reliably. In particular this is not oom_kill-friendly. This patch tries to fix the 1st problem, and makes the preparation for the next changes. We add the new SIGNAL_GROUP_COREDUMP flag set by zap_threads() to indicate that this process dumps the core. prepare_signal() checks this flag and nacks any signal except SIGKILL. Note that this check tries to be conservative, in the long term we should probably treat the SIGNAL_GROUP_EXIT case equally but this needs more discussion. See marc.info/?l=linux-kernel&m=120508897917439 Notes: - recalc_sigpending() doesn't check SIGNAL_GROUP_COREDUMP. The patch assumes that dump_write/etc paths should never call it, but we can change it as well. - There is another source of TIF_SIGPENDING, freezer. This will be addressed separately. Signed-off-by: Oleg Nesterov <oleg@redhat.com> Tested-by: Mandeep Singh Baines <msb@chromium.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Neil Horman <nhorman@redhat.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Roland McGrath <roland@hack.frob.com> Cc: Tejun Heo <tj@kernel.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 22:28:10 +00:00
clear_tsk_thread_flag(tsk, TIF_SIGPENDING);
coredump: Limit coredumps to a single thread group Today when a signal is delivered with a handler of SIG_DFL whose default behavior is to generate a core dump not only that process but every process that shares the mm is killed. In the case of vfork this looks like a real world problem. Consider the following well defined sequence. if (vfork() == 0) { execve(...); _exit(EXIT_FAILURE); } If a signal that generates a core dump is received after vfork but before the execve changes the mm the process that called vfork will also be killed (as the mm is shared). Similarly if the execve fails after the point of no return the kernel delivers SIGSEGV which will kill both the exec'ing process and because the mm is shared the process that called vfork as well. As far as I can tell this behavior is a violation of people's reasonable expectations, POSIX, and is unnecessarily fragile when the system is low on memory. Solve this by making a userspace visible change to only kill a single process/thread group. This is possible because Jann Horn recently modified[1] the coredump code so that the mm can safely be modified while the coredump is happening. With LinuxThreads long gone I don't expect anyone to have a notice this behavior change in practice. To accomplish this move the core_state pointer from mm_struct to signal_struct, which allows different thread groups to coredump simultatenously. In zap_threads remove the work to kill anything except for the current thread group. v2: Remove core_state from the VM_BUG_ON_MM print to fix compile failure when CONFIG_DEBUG_VM is enabled. Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> [1] a07279c9a8cd ("binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot") Fixes: d89f3847def4 ("[PATCH] thread-aware coredumps, 2.5.43-C3") History-tree: git://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git Link: https://lkml.kernel.org/r/87y27mvnke.fsf@disp2133 Link: https://lkml.kernel.org/r/20211007144701.67592574@canb.auug.org.au Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2021-09-22 16:24:02 +00:00
tsk->flags |= PF_DUMPCORE;
atomic_set(&core_state->nr_threads, nr);
}
spin_unlock_irq(&tsk->sighand->siglock);
return nr;
}
static int coredump_wait(int exit_code, struct core_state *core_state)
{
struct task_struct *tsk = current;
int core_waiters = -EBUSY;
init_completion(&core_state->startup);
core_state->dumper.task = tsk;
core_state->dumper.next = NULL;
coredump: Limit coredumps to a single thread group Today when a signal is delivered with a handler of SIG_DFL whose default behavior is to generate a core dump not only that process but every process that shares the mm is killed. In the case of vfork this looks like a real world problem. Consider the following well defined sequence. if (vfork() == 0) { execve(...); _exit(EXIT_FAILURE); } If a signal that generates a core dump is received after vfork but before the execve changes the mm the process that called vfork will also be killed (as the mm is shared). Similarly if the execve fails after the point of no return the kernel delivers SIGSEGV which will kill both the exec'ing process and because the mm is shared the process that called vfork as well. As far as I can tell this behavior is a violation of people's reasonable expectations, POSIX, and is unnecessarily fragile when the system is low on memory. Solve this by making a userspace visible change to only kill a single process/thread group. This is possible because Jann Horn recently modified[1] the coredump code so that the mm can safely be modified while the coredump is happening. With LinuxThreads long gone I don't expect anyone to have a notice this behavior change in practice. To accomplish this move the core_state pointer from mm_struct to signal_struct, which allows different thread groups to coredump simultatenously. In zap_threads remove the work to kill anything except for the current thread group. v2: Remove core_state from the VM_BUG_ON_MM print to fix compile failure when CONFIG_DEBUG_VM is enabled. Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> [1] a07279c9a8cd ("binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot") Fixes: d89f3847def4 ("[PATCH] thread-aware coredumps, 2.5.43-C3") History-tree: git://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git Link: https://lkml.kernel.org/r/87y27mvnke.fsf@disp2133 Link: https://lkml.kernel.org/r/20211007144701.67592574@canb.auug.org.au Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2021-09-22 16:24:02 +00:00
core_waiters = zap_threads(tsk, core_state, exit_code);
if (core_waiters > 0) {
struct core_thread *ptr;
freezer,sched: Rewrite core freezer logic Rewrite the core freezer to behave better wrt thawing and be simpler in general. By replacing PF_FROZEN with TASK_FROZEN, a special block state, it is ensured frozen tasks stay frozen until thawed and don't randomly wake up early, as is currently possible. As such, it does away with PF_FROZEN and PF_FREEZER_SKIP, freeing up two PF_flags (yay!). Specifically; the current scheme works a little like: freezer_do_not_count(); schedule(); freezer_count(); And either the task is blocked, or it lands in try_to_freezer() through freezer_count(). Now, when it is blocked, the freezer considers it frozen and continues. However, on thawing, once pm_freezing is cleared, freezer_count() stops working, and any random/spurious wakeup will let a task run before its time. That is, thawing tries to thaw things in explicit order; kernel threads and workqueues before doing bringing SMP back before userspace etc.. However due to the above mentioned races it is entirely possible for userspace tasks to thaw (by accident) before SMP is back. This can be a fatal problem in asymmetric ISA architectures (eg ARMv9) where the userspace task requires a special CPU to run. As said; replace this with a special task state TASK_FROZEN and add the following state transitions: TASK_FREEZABLE -> TASK_FROZEN __TASK_STOPPED -> TASK_FROZEN __TASK_TRACED -> TASK_FROZEN The new TASK_FREEZABLE can be set on any state part of TASK_NORMAL (IOW. TASK_INTERRUPTIBLE and TASK_UNINTERRUPTIBLE) -- any such state is already required to deal with spurious wakeups and the freezer causes one such when thawing the task (since the original state is lost). The special __TASK_{STOPPED,TRACED} states *can* be restored since their canonical state is in ->jobctl. With this, frozen tasks need an explicit TASK_FROZEN wakeup and are free of undue (early / spurious) wakeups. Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Reviewed-by: Ingo Molnar <mingo@kernel.org> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Link: https://lore.kernel.org/r/20220822114649.055452969@infradead.org
2022-08-22 11:18:22 +00:00
wait_for_completion_state(&core_state->startup,
TASK_UNINTERRUPTIBLE|TASK_FREEZABLE);
/*
* Wait for all the threads to become inactive, so that
* all the thread context (extended register state, like
* fpu etc) gets copied to the memory.
*/
ptr = core_state->dumper.next;
while (ptr != NULL) {
wait_task_inactive(ptr->task, TASK_ANY);
ptr = ptr->next;
}
}
return core_waiters;
}
coredump: Limit coredumps to a single thread group Today when a signal is delivered with a handler of SIG_DFL whose default behavior is to generate a core dump not only that process but every process that shares the mm is killed. In the case of vfork this looks like a real world problem. Consider the following well defined sequence. if (vfork() == 0) { execve(...); _exit(EXIT_FAILURE); } If a signal that generates a core dump is received after vfork but before the execve changes the mm the process that called vfork will also be killed (as the mm is shared). Similarly if the execve fails after the point of no return the kernel delivers SIGSEGV which will kill both the exec'ing process and because the mm is shared the process that called vfork as well. As far as I can tell this behavior is a violation of people's reasonable expectations, POSIX, and is unnecessarily fragile when the system is low on memory. Solve this by making a userspace visible change to only kill a single process/thread group. This is possible because Jann Horn recently modified[1] the coredump code so that the mm can safely be modified while the coredump is happening. With LinuxThreads long gone I don't expect anyone to have a notice this behavior change in practice. To accomplish this move the core_state pointer from mm_struct to signal_struct, which allows different thread groups to coredump simultatenously. In zap_threads remove the work to kill anything except for the current thread group. v2: Remove core_state from the VM_BUG_ON_MM print to fix compile failure when CONFIG_DEBUG_VM is enabled. Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> [1] a07279c9a8cd ("binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot") Fixes: d89f3847def4 ("[PATCH] thread-aware coredumps, 2.5.43-C3") History-tree: git://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git Link: https://lkml.kernel.org/r/87y27mvnke.fsf@disp2133 Link: https://lkml.kernel.org/r/20211007144701.67592574@canb.auug.org.au Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2021-09-22 16:24:02 +00:00
static void coredump_finish(bool core_dumped)
{
struct core_thread *curr, *next;
struct task_struct *task;
coredump: ensure that SIGKILL always kills the dumping thread prepare_signal() blesses SIGKILL sent to the dumping process but this signal can be "lost" anyway. The problems is, complete_signal() sees SIGNAL_GROUP_EXIT and skips the "kill them all" logic. And even if the dumping process is single-threaded (so the target is always "correct"), the group-wide SIGKILL is not recorded in task->pending and thus __fatal_signal_pending() won't be true. A multi-threaded case has even more problems. And even ignoring all technical details, SIGNAL_GROUP_EXIT doesn't look right to me. This coredumping process is not exiting yet, it can do a lot of work dumping the core. With this patch the dumping process doesn't have SIGNAL_GROUP_EXIT, we set signal->group_exit_task instead. This makes signal_group_exit() true and thus this should equally close the races with exit/exec/stop but allows to kill the dumping thread reliably. Notes: - It is not clear what should we do with ->group_exit_code if the dumper was killed, see the next change. - we need more (hopefully straightforward) changes to ensure that SIGKILL actually interrupts the coredump. Basically we need to check __fatal_signal_pending() in dump_write() and dump_seek(). Signed-off-by: Oleg Nesterov <oleg@redhat.com> Tested-by: Mandeep Singh Baines <msb@chromium.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Neil Horman <nhorman@redhat.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Roland McGrath <roland@hack.frob.com> Cc: Tejun Heo <tj@kernel.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 22:28:12 +00:00
spin_lock_irq(&current->sighand->siglock);
if (core_dumped && !__fatal_signal_pending(current))
current->signal->group_exit_code |= 0x80;
coredump: Limit coredumps to a single thread group Today when a signal is delivered with a handler of SIG_DFL whose default behavior is to generate a core dump not only that process but every process that shares the mm is killed. In the case of vfork this looks like a real world problem. Consider the following well defined sequence. if (vfork() == 0) { execve(...); _exit(EXIT_FAILURE); } If a signal that generates a core dump is received after vfork but before the execve changes the mm the process that called vfork will also be killed (as the mm is shared). Similarly if the execve fails after the point of no return the kernel delivers SIGSEGV which will kill both the exec'ing process and because the mm is shared the process that called vfork as well. As far as I can tell this behavior is a violation of people's reasonable expectations, POSIX, and is unnecessarily fragile when the system is low on memory. Solve this by making a userspace visible change to only kill a single process/thread group. This is possible because Jann Horn recently modified[1] the coredump code so that the mm can safely be modified while the coredump is happening. With LinuxThreads long gone I don't expect anyone to have a notice this behavior change in practice. To accomplish this move the core_state pointer from mm_struct to signal_struct, which allows different thread groups to coredump simultatenously. In zap_threads remove the work to kill anything except for the current thread group. v2: Remove core_state from the VM_BUG_ON_MM print to fix compile failure when CONFIG_DEBUG_VM is enabled. Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> [1] a07279c9a8cd ("binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot") Fixes: d89f3847def4 ("[PATCH] thread-aware coredumps, 2.5.43-C3") History-tree: git://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git Link: https://lkml.kernel.org/r/87y27mvnke.fsf@disp2133 Link: https://lkml.kernel.org/r/20211007144701.67592574@canb.auug.org.au Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2021-09-22 16:24:02 +00:00
next = current->signal->core_state->dumper.next;
current->signal->core_state = NULL;
coredump: ensure that SIGKILL always kills the dumping thread prepare_signal() blesses SIGKILL sent to the dumping process but this signal can be "lost" anyway. The problems is, complete_signal() sees SIGNAL_GROUP_EXIT and skips the "kill them all" logic. And even if the dumping process is single-threaded (so the target is always "correct"), the group-wide SIGKILL is not recorded in task->pending and thus __fatal_signal_pending() won't be true. A multi-threaded case has even more problems. And even ignoring all technical details, SIGNAL_GROUP_EXIT doesn't look right to me. This coredumping process is not exiting yet, it can do a lot of work dumping the core. With this patch the dumping process doesn't have SIGNAL_GROUP_EXIT, we set signal->group_exit_task instead. This makes signal_group_exit() true and thus this should equally close the races with exit/exec/stop but allows to kill the dumping thread reliably. Notes: - It is not clear what should we do with ->group_exit_code if the dumper was killed, see the next change. - we need more (hopefully straightforward) changes to ensure that SIGKILL actually interrupts the coredump. Basically we need to check __fatal_signal_pending() in dump_write() and dump_seek(). Signed-off-by: Oleg Nesterov <oleg@redhat.com> Tested-by: Mandeep Singh Baines <msb@chromium.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Neil Horman <nhorman@redhat.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Roland McGrath <roland@hack.frob.com> Cc: Tejun Heo <tj@kernel.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 22:28:12 +00:00
spin_unlock_irq(&current->sighand->siglock);
while ((curr = next) != NULL) {
next = curr->next;
task = curr->task;
/*
coredump: Don't perform any cleanups before dumping core Rename coredump_exit_mm to coredump_task_exit and call it from do_exit before PTRACE_EVENT_EXIT, and before any cleanup work for a task happens. This ensures that an accurate copy of the process can be captured in the coredump as no cleanup for the process happens before the coredump completes. This also ensures that PTRACE_EVENT_EXIT will not be visited by any thread until the coredump is complete. Add a new flag PF_POSTCOREDUMP so that tasks that have passed through coredump_task_exit can be recognized and ignored in zap_process. Now that all of the coredumping happens before exit_mm remove code to test for a coredump in progress from mm_release. Replace "may_ptrace_stop()" with a simple test of "current->ptrace". The other tests in may_ptrace_stop all concern avoiding stopping during a coredump. These tests are no longer necessary as it is now guaranteed that fatal_signal_pending will be set if the code enters ptrace_stop during a coredump. The code in ptrace_stop is guaranteed not to stop if fatal_signal_pending returns true. Until this change "ptrace_event(PTRACE_EVENT_EXIT)" could call ptrace_stop without fatal_signal_pending being true, as signals are dequeued in get_signal before calling do_exit. This is no longer an issue as "ptrace_event(PTRACE_EVENT_EXIT)" is no longer reached until after the coredump completes. Link: https://lkml.kernel.org/r/874kaax26c.fsf@disp2133 Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2021-09-01 16:33:50 +00:00
* see coredump_task_exit(), curr->task must not see
* ->task == NULL before we read ->next.
*/
smp_mb();
curr->task = NULL;
wake_up_process(task);
}
}
coredump: introduce dump_interrupted() By discussion with Mandeep. Change dump_write(), dump_seek() and do_coredump() to check signal_pending() and abort if it is true. dump_seek() does this only before f_op->llseek(), otherwise it relies on dump_write(). We need this change to ensure that the coredump won't delay suspend, and to ensure it reacts to SIGKILL "quickly enough", a core dump can take a lot of time. In particular this can help oom-killer. We add the new trivial helper, dump_interrupted() to add the comments and to simplify the potential freezer changes. Perhaps it will have more callers. Ideally it should do try_to_freeze() but then we need the unpleasant changes in dump_write() and wait_for_dump_helpers(). It is not trivial to change dump_write() to restart if f_op->write() fails because of freezing(). We need to handle the short writes, we need to clear TIF_SIGPENDING (and we can't rely on recalc_sigpending() unless we change it to check PF_DUMPCORE). And if the buggy f_op->write() sets TIF_SIGPENDING we can not distinguish this case from the race with freeze_task() + __thaw_task(). So we simply accept the fact that the freezer can truncate a core-dump but at least you can reliably suspend. Hopefully we can tolerate this unlikely case and the necessary complications doesn't worth a trouble. But if we decide to make the coredumping freezable later we can do this on top of this change. Signed-off-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Mandeep Singh Baines <msb@chromium.org> Cc: Neil Horman <nhorman@redhat.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Tejun Heo <tj@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 22:28:15 +00:00
static bool dump_interrupted(void)
{
/*
* SIGKILL or freezing() interrupt the coredumping. Perhaps we
* can do try_to_freeze() and check __fatal_signal_pending(),
* but then we need to teach dump_write() to restart and clear
* TIF_SIGPENDING.
*/
coredump: Limit what can interrupt coredumps Olivier Langlois has been struggling with coredumps being incompletely written in processes using io_uring. Olivier Langlois <olivier@trillion01.com> writes: > io_uring is a big user of task_work and any event that io_uring made a > task waiting for that occurs during the core dump generation will > generate a TIF_NOTIFY_SIGNAL. > > Here are the detailed steps of the problem: > 1. io_uring calls vfs_poll() to install a task to a file wait queue > with io_async_wake() as the wakeup function cb from io_arm_poll_handler() > 2. wakeup function ends up calling task_work_add() with TWA_SIGNAL > 3. task_work_add() sets the TIF_NOTIFY_SIGNAL bit by calling > set_notify_signal() The coredump code deliberately supports being interrupted by SIGKILL, and depends upon prepare_signal to filter out all other signals. Now that signal_pending includes wake ups for TIF_NOTIFY_SIGNAL this hack in dump_emitted by the coredump code no longer works. Make the coredump code more robust by explicitly testing for all of the wakeup conditions the coredump code supports. This prevents new wakeup conditions from breaking the coredump code, as well as fixing the current issue. The filesystem code that the coredump code uses already limits itself to only aborting on fatal_signal_pending. So it should not develop surprising wake-up reasons either. v2: Don't remove the now unnecessary code in prepare_signal. Cc: stable@vger.kernel.org Fixes: 12db8b690010 ("entry: Add support for TIF_NOTIFY_SIGNAL") Reported-by: Olivier Langlois <olivier@trillion01.com> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-06-10 20:11:11 +00:00
return fatal_signal_pending(current) || freezing(current);
coredump: introduce dump_interrupted() By discussion with Mandeep. Change dump_write(), dump_seek() and do_coredump() to check signal_pending() and abort if it is true. dump_seek() does this only before f_op->llseek(), otherwise it relies on dump_write(). We need this change to ensure that the coredump won't delay suspend, and to ensure it reacts to SIGKILL "quickly enough", a core dump can take a lot of time. In particular this can help oom-killer. We add the new trivial helper, dump_interrupted() to add the comments and to simplify the potential freezer changes. Perhaps it will have more callers. Ideally it should do try_to_freeze() but then we need the unpleasant changes in dump_write() and wait_for_dump_helpers(). It is not trivial to change dump_write() to restart if f_op->write() fails because of freezing(). We need to handle the short writes, we need to clear TIF_SIGPENDING (and we can't rely on recalc_sigpending() unless we change it to check PF_DUMPCORE). And if the buggy f_op->write() sets TIF_SIGPENDING we can not distinguish this case from the race with freeze_task() + __thaw_task(). So we simply accept the fact that the freezer can truncate a core-dump but at least you can reliably suspend. Hopefully we can tolerate this unlikely case and the necessary complications doesn't worth a trouble. But if we decide to make the coredumping freezable later we can do this on top of this change. Signed-off-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Mandeep Singh Baines <msb@chromium.org> Cc: Neil Horman <nhorman@redhat.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Tejun Heo <tj@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 22:28:15 +00:00
}
static void wait_for_dump_helpers(struct file *file)
{
struct pipe_inode_info *pipe = file->private_data;
pipe_lock(pipe);
pipe->readers++;
pipe->writers--;
pipe: use exclusive waits when reading or writing This makes the pipe code use separate wait-queues and exclusive waiting for readers and writers, avoiding a nasty thundering herd problem when there are lots of readers waiting for data on a pipe (or, less commonly, lots of writers waiting for a pipe to have space). While this isn't a common occurrence in the traditional "use a pipe as a data transport" case, where you typically only have a single reader and a single writer process, there is one common special case: using a pipe as a source of "locking tokens" rather than for data communication. In particular, the GNU make jobserver code ends up using a pipe as a way to limit parallelism, where each job consumes a token by reading a byte from the jobserver pipe, and releases the token by writing a byte back to the pipe. This pattern is fairly traditional on Unix, and works very well, but will waste a lot of time waking up a lot of processes when only a single reader needs to be woken up when a writer releases a new token. A simplified test-case of just this pipe interaction is to create 64 processes, and then pass a single token around between them (this test-case also intentionally passes another token that gets ignored to test the "wake up next" logic too, in case anybody wonders about it): #include <unistd.h> int main(int argc, char **argv) { int fd[2], counters[2]; pipe(fd); counters[0] = 0; counters[1] = -1; write(fd[1], counters, sizeof(counters)); /* 64 processes */ fork(); fork(); fork(); fork(); fork(); fork(); do { int i; read(fd[0], &i, sizeof(i)); if (i < 0) continue; counters[0] = i+1; write(fd[1], counters, (1+(i & 1)) *sizeof(int)); } while (counters[0] < 1000000); return 0; } and in a perfect world, passing that token around should only cause one context switch per transfer, when the writer of a token causes a directed wakeup of just a single reader. But with the "writer wakes all readers" model we traditionally had, on my test box the above case causes more than an order of magnitude more scheduling: instead of the expected ~1M context switches, "perf stat" shows 231,852.37 msec task-clock # 15.857 CPUs utilized 11,250,961 context-switches # 0.049 M/sec 616,304 cpu-migrations # 0.003 M/sec 1,648 page-faults # 0.007 K/sec 1,097,903,998,514 cycles # 4.735 GHz 120,781,778,352 instructions # 0.11 insn per cycle 27,997,056,043 branches # 120.754 M/sec 283,581,233 branch-misses # 1.01% of all branches 14.621273891 seconds time elapsed 0.018243000 seconds user 3.611468000 seconds sys before this commit. After this commit, I get 5,229.55 msec task-clock # 3.072 CPUs utilized 1,212,233 context-switches # 0.232 M/sec 103,951 cpu-migrations # 0.020 M/sec 1,328 page-faults # 0.254 K/sec 21,307,456,166 cycles # 4.074 GHz 12,947,819,999 instructions # 0.61 insn per cycle 2,881,985,678 branches # 551.096 M/sec 64,267,015 branch-misses # 2.23% of all branches 1.702148350 seconds time elapsed 0.004868000 seconds user 0.110786000 seconds sys instead. Much better. [ Note! This kernel improvement seems to be very good at triggering a race condition in the make jobserver (in GNU make 4.2.1) for me. It's a long known bug that was fixed back in June 2017 by GNU make commit b552b0525198 ("[SV 51159] Use a non-blocking read with pselect to avoid hangs."). But there wasn't a new release of GNU make until 4.3 on Jan 19 2020, so a number of distributions may still have the buggy version. Some have backported the fix to their 4.2.1 release, though, and even without the fix it's quite timing-dependent whether the bug actually is hit. ] Josh Triplett says: "I've been hammering on your pipe fix patch (switching to exclusive wait queues) for a month or so, on several different systems, and I've run into no issues with it. The patch *substantially* improves parallel build times on large (~100 CPU) systems, both with parallel make and with other things that use make's pipe-based jobserver. All current distributions (including stable and long-term stable distributions) have versions of GNU make that no longer have the jobserver bug" Tested-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-12-09 17:48:27 +00:00
wake_up_interruptible_sync(&pipe->rd_wait);
kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
pipe_unlock(pipe);
/*
* We actually want wait_event_freezable() but then we need
* to clear TIF_SIGPENDING and improve dump_interrupted().
*/
pipe: use exclusive waits when reading or writing This makes the pipe code use separate wait-queues and exclusive waiting for readers and writers, avoiding a nasty thundering herd problem when there are lots of readers waiting for data on a pipe (or, less commonly, lots of writers waiting for a pipe to have space). While this isn't a common occurrence in the traditional "use a pipe as a data transport" case, where you typically only have a single reader and a single writer process, there is one common special case: using a pipe as a source of "locking tokens" rather than for data communication. In particular, the GNU make jobserver code ends up using a pipe as a way to limit parallelism, where each job consumes a token by reading a byte from the jobserver pipe, and releases the token by writing a byte back to the pipe. This pattern is fairly traditional on Unix, and works very well, but will waste a lot of time waking up a lot of processes when only a single reader needs to be woken up when a writer releases a new token. A simplified test-case of just this pipe interaction is to create 64 processes, and then pass a single token around between them (this test-case also intentionally passes another token that gets ignored to test the "wake up next" logic too, in case anybody wonders about it): #include <unistd.h> int main(int argc, char **argv) { int fd[2], counters[2]; pipe(fd); counters[0] = 0; counters[1] = -1; write(fd[1], counters, sizeof(counters)); /* 64 processes */ fork(); fork(); fork(); fork(); fork(); fork(); do { int i; read(fd[0], &i, sizeof(i)); if (i < 0) continue; counters[0] = i+1; write(fd[1], counters, (1+(i & 1)) *sizeof(int)); } while (counters[0] < 1000000); return 0; } and in a perfect world, passing that token around should only cause one context switch per transfer, when the writer of a token causes a directed wakeup of just a single reader. But with the "writer wakes all readers" model we traditionally had, on my test box the above case causes more than an order of magnitude more scheduling: instead of the expected ~1M context switches, "perf stat" shows 231,852.37 msec task-clock # 15.857 CPUs utilized 11,250,961 context-switches # 0.049 M/sec 616,304 cpu-migrations # 0.003 M/sec 1,648 page-faults # 0.007 K/sec 1,097,903,998,514 cycles # 4.735 GHz 120,781,778,352 instructions # 0.11 insn per cycle 27,997,056,043 branches # 120.754 M/sec 283,581,233 branch-misses # 1.01% of all branches 14.621273891 seconds time elapsed 0.018243000 seconds user 3.611468000 seconds sys before this commit. After this commit, I get 5,229.55 msec task-clock # 3.072 CPUs utilized 1,212,233 context-switches # 0.232 M/sec 103,951 cpu-migrations # 0.020 M/sec 1,328 page-faults # 0.254 K/sec 21,307,456,166 cycles # 4.074 GHz 12,947,819,999 instructions # 0.61 insn per cycle 2,881,985,678 branches # 551.096 M/sec 64,267,015 branch-misses # 2.23% of all branches 1.702148350 seconds time elapsed 0.004868000 seconds user 0.110786000 seconds sys instead. Much better. [ Note! This kernel improvement seems to be very good at triggering a race condition in the make jobserver (in GNU make 4.2.1) for me. It's a long known bug that was fixed back in June 2017 by GNU make commit b552b0525198 ("[SV 51159] Use a non-blocking read with pselect to avoid hangs."). But there wasn't a new release of GNU make until 4.3 on Jan 19 2020, so a number of distributions may still have the buggy version. Some have backported the fix to their 4.2.1 release, though, and even without the fix it's quite timing-dependent whether the bug actually is hit. ] Josh Triplett says: "I've been hammering on your pipe fix patch (switching to exclusive wait queues) for a month or so, on several different systems, and I've run into no issues with it. The patch *substantially* improves parallel build times on large (~100 CPU) systems, both with parallel make and with other things that use make's pipe-based jobserver. All current distributions (including stable and long-term stable distributions) have versions of GNU make that no longer have the jobserver bug" Tested-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-12-09 17:48:27 +00:00
wait_event_interruptible(pipe->rd_wait, pipe->readers == 1);
pipe_lock(pipe);
pipe->readers--;
pipe->writers++;
pipe_unlock(pipe);
}
/*
* umh_pipe_setup
* helper function to customize the process used
* to collect the core in userspace. Specifically
* it sets up a pipe and installs it as fd 0 (stdin)
* for the process. Returns 0 on success, or
* PTR_ERR on failure.
* Note that it also sets the core limit to 1. This
* is a special value that we use to trap recursive
* core dumps
*/
static int umh_pipe_setup(struct subprocess_info *info, struct cred *new)
{
struct file *files[2];
struct coredump_params *cp = (struct coredump_params *)info->data;
int err = create_pipe_files(files, 0);
if (err)
return err;
cp->file = files[1];
err = replace_fd(0, files[0], 0);
fput(files[0]);
/* and disallow core files too */
current->signal->rlim[RLIMIT_CORE] = (struct rlimit){1, 1};
return err;
}
void do_coredump(const kernel_siginfo_t *siginfo)
{
struct core_state core_state;
struct core_name cn;
struct mm_struct *mm = current->mm;
struct linux_binfmt * binfmt;
const struct cred *old_cred;
struct cred *cred;
int retval = 0;
int ispipe;
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
size_t *argv = NULL;
int argc = 0;
fs: if a coredump already exists, unlink and recreate with O_EXCL It was possible for an attacking user to trick root (or another user) into writing his coredumps into an attacker-readable, pre-existing file using rename() or link(), causing the disclosure of secret data from the victim process' virtual memory. Depending on the configuration, it was also possible to trick root into overwriting system files with coredumps. Fix that issue by never writing coredumps into existing files. Requirements for the attack: - The attack only applies if the victim's process has a nonzero RLIMIT_CORE and is dumpable. - The attacker can trick the victim into coredumping into an attacker-writable directory D, either because the core_pattern is relative and the victim's cwd is attacker-writable or because an absolute core_pattern pointing to a world-writable directory is used. - The attacker has one of these: A: on a system with protected_hardlinks=0: execute access to a folder containing a victim-owned, attacker-readable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (In practice, there are lots of files that fulfill this condition, e.g. entries in Debian's /var/lib/dpkg/info/.) This does not apply to most Linux systems because most distros set protected_hardlinks=1. B: on a system with protected_hardlinks=1: execute access to a folder containing a victim-owned, attacker-readable and attacker-writable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (This seems to be uncommon.) C: on any system, independent of protected_hardlinks: write access to a non-sticky folder containing a victim-owned, attacker-readable file on the same partition as D (This seems to be uncommon.) The basic idea is that the attacker moves the victim-owned file to where he expects the victim process to dump its core. The victim process dumps its core into the existing file, and the attacker reads the coredump from it. If the attacker can't move the file because he does not have write access to the containing directory, he can instead link the file to a directory he controls, then wait for the original link to the file to be deleted (because the kernel checks that the link count of the corefile is 1). A less reliable variant that requires D to be non-sticky works with link() and does not require deletion of the original link: link() the file into D, but then unlink() it directly before the kernel performs the link count check. On systems with protected_hardlinks=0, this variant allows an attacker to not only gain information from coredumps, but also clobber existing, victim-writable files with coredumps. (This could theoretically lead to a privilege escalation.) Signed-off-by: Jann Horn <jann@thejh.net> Cc: Kees Cook <keescook@chromium.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-09 22:38:28 +00:00
/* require nonrelative corefile path and be extra careful */
bool need_suid_safe = false;
bool core_dumped = false;
static atomic_t core_dump_count = ATOMIC_INIT(0);
struct coredump_params cprm = {
.siginfo = siginfo,
.limit = rlimit(RLIMIT_CORE),
/*
* We must use the same mm->flags while dumping core to avoid
* inconsistency of bit flags, since this flag is not protected
* by any locks.
*/
.mm_flags = mm->flags,
.vma_meta = NULL,
.cpu = raw_smp_processor_id(),
};
audit_core_dumps(siginfo->si_signo);
binfmt = mm->binfmt;
if (!binfmt || !binfmt->core_dump)
goto fail;
if (!__get_dumpable(cprm.mm_flags))
goto fail;
cred = prepare_creds();
if (!cred)
goto fail;
/*
* We cannot trust fsuid as being the "true" uid of the process
* nor do we know its entire history. We only know it was tainted
* so we dump it as root in mode 2, and only into a controlled
* environment (pipe handler or fully qualified path).
*/
if (__get_dumpable(cprm.mm_flags) == SUID_DUMP_ROOT) {
/* Setuid core dump mode */
cred->fsuid = GLOBAL_ROOT_UID; /* Dump root private */
fs: if a coredump already exists, unlink and recreate with O_EXCL It was possible for an attacking user to trick root (or another user) into writing his coredumps into an attacker-readable, pre-existing file using rename() or link(), causing the disclosure of secret data from the victim process' virtual memory. Depending on the configuration, it was also possible to trick root into overwriting system files with coredumps. Fix that issue by never writing coredumps into existing files. Requirements for the attack: - The attack only applies if the victim's process has a nonzero RLIMIT_CORE and is dumpable. - The attacker can trick the victim into coredumping into an attacker-writable directory D, either because the core_pattern is relative and the victim's cwd is attacker-writable or because an absolute core_pattern pointing to a world-writable directory is used. - The attacker has one of these: A: on a system with protected_hardlinks=0: execute access to a folder containing a victim-owned, attacker-readable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (In practice, there are lots of files that fulfill this condition, e.g. entries in Debian's /var/lib/dpkg/info/.) This does not apply to most Linux systems because most distros set protected_hardlinks=1. B: on a system with protected_hardlinks=1: execute access to a folder containing a victim-owned, attacker-readable and attacker-writable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (This seems to be uncommon.) C: on any system, independent of protected_hardlinks: write access to a non-sticky folder containing a victim-owned, attacker-readable file on the same partition as D (This seems to be uncommon.) The basic idea is that the attacker moves the victim-owned file to where he expects the victim process to dump its core. The victim process dumps its core into the existing file, and the attacker reads the coredump from it. If the attacker can't move the file because he does not have write access to the containing directory, he can instead link the file to a directory he controls, then wait for the original link to the file to be deleted (because the kernel checks that the link count of the corefile is 1). A less reliable variant that requires D to be non-sticky works with link() and does not require deletion of the original link: link() the file into D, but then unlink() it directly before the kernel performs the link count check. On systems with protected_hardlinks=0, this variant allows an attacker to not only gain information from coredumps, but also clobber existing, victim-writable files with coredumps. (This could theoretically lead to a privilege escalation.) Signed-off-by: Jann Horn <jann@thejh.net> Cc: Kees Cook <keescook@chromium.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-09 22:38:28 +00:00
need_suid_safe = true;
}
retval = coredump_wait(siginfo->si_signo, &core_state);
if (retval < 0)
goto fail_creds;
old_cred = override_creds(cred);
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
ispipe = format_corename(&cn, &cprm, &argv, &argc);
if (ispipe) {
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
int argi;
int dump_count;
char **helper_argv;
struct subprocess_info *sub_info;
if (ispipe < 0) {
printk(KERN_WARNING "format_corename failed\n");
printk(KERN_WARNING "Aborting core\n");
goto fail_unlock;
}
if (cprm.limit == 1) {
/* See umh_pipe_setup() which sets RLIMIT_CORE = 1.
*
* Normally core limits are irrelevant to pipes, since
* we're not writing to the file system, but we use
* cprm.limit of 1 here as a special value, this is a
* consistent way to catch recursive crashes.
* We can still crash if the core_pattern binary sets
* RLIM_CORE = !1, but it runs as root, and can do
* lots of stupid things.
*
* Note that we use task_tgid_vnr here to grab the pid
* of the process group leader. That way we get the
* right pid if a thread in a multi-threaded
* core_pattern process dies.
*/
printk(KERN_WARNING
"Process %d(%s) has RLIMIT_CORE set to 1\n",
task_tgid_vnr(current), current->comm);
printk(KERN_WARNING "Aborting core\n");
goto fail_unlock;
}
cprm.limit = RLIM_INFINITY;
dump_count = atomic_inc_return(&core_dump_count);
if (core_pipe_limit && (core_pipe_limit < dump_count)) {
printk(KERN_WARNING "Pid %d(%s) over core_pipe_limit\n",
task_tgid_vnr(current), current->comm);
printk(KERN_WARNING "Skipping core dump\n");
goto fail_dropcount;
}
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
helper_argv = kmalloc_array(argc + 1, sizeof(*helper_argv),
GFP_KERNEL);
if (!helper_argv) {
printk(KERN_WARNING "%s failed to allocate memory\n",
__func__);
goto fail_dropcount;
}
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
for (argi = 0; argi < argc; argi++)
helper_argv[argi] = cn.corename + argv[argi];
helper_argv[argi] = NULL;
retval = -ENOMEM;
sub_info = call_usermodehelper_setup(helper_argv[0],
helper_argv, NULL, GFP_KERNEL,
umh_pipe_setup, NULL, &cprm);
if (sub_info)
retval = call_usermodehelper_exec(sub_info,
UMH_WAIT_EXEC);
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
kfree(helper_argv);
if (retval) {
printk(KERN_INFO "Core dump to |%s pipe failed\n",
cn.corename);
goto close_fail;
}
} else {
struct mnt_idmap *idmap;
struct inode *inode;
coredump: require O_WRONLY instead of O_RDWR The motivation for this patch has been to enable using a stricter apparmor profile to prevent programs from reading any coredump in the system. However, this became something else. The following details are based on Christian's and Linus' archeology into the history of the number "2" in the coredump handling code. To make sure we're not accidently introducing some subtle behavioral change into the coredump code we set out on a voyage into the depths of history.git to figure out why this was O_RDWR in the first place. Coredump handling was introduced over 30 years ago in commit ddc733f452e0 ("[PATCH] Linux-0.97 (August 1, 1992)"). The original code used O_WRONLY: open_namei("core",O_CREAT | O_WRONLY | O_TRUNC,0600,&inode,NULL) However, this changed in 1993 and starting with commit 9cb9f18b5d26 ("[PATCH] Linux-0.99.10 (June 7, 1993)") the coredump code suddenly used the constant "2": open_namei("core",O_CREAT | 2 | O_TRUNC,0600,&inode,NULL) This was curious as in the same commit the kernel switched from constants to proper defines in other places such as KERNEL_DS and USER_DS and O_RDWR did already exist. So why was "2" used? It turns out that open_namei() - an early version of what later turned into filp_open() - didn't accept O_RDWR. A semantic quirk of the open() uapi is the definition of the O_RDONLY flag. It would seem natural to define: #define O_RDWR (O_RDONLY | O_WRONLY) but that isn't possible because: #define O_RDONLY 0 This makes O_RDONLY effectively meaningless when passed to the kernel. In other words, there has never been a way - until O_PATH at least - to open a file without any permission; O_RDONLY was always implied on the uapi side while the kernel does in fact allow opening files without permissions. The trouble comes when trying to map the uapi flags onto the corresponding file mode flags FMODE_{READ,WRITE}. This mapping still happens today and is causing issues to this day (We ran into this during additions for openat2() for example.). So the special value "3" was used to indicate that the file was opened for special access: f->f_flags = flag = flags; f->f_mode = (flag+1) & O_ACCMODE; if (f->f_mode) flag++; This allowed the file mode to be set to FMODE_READ | FMODE_WRITE mapping the O_{RDONLY,WRONLY,RDWR} flags into the FMODE_{READ,WRITE} flags. The special access then required read-write permissions and 0 was used to access symlinks. But back when ddc733f452e0 ("[PATCH] Linux-0.97 (August 1, 1992)") added coredump handling open_namei() took the FMODE_{READ,WRITE} flags as an argument. So the coredump handling introduced in ddc733f452e0 ("[PATCH] Linux-0.97 (August 1, 1992)") was buggy because O_WRONLY shouldn't have been passed. Since O_WRONLY is 1 but open_namei() took FMODE_{READ,WRITE} it was passed FMODE_READ on accident. So 9cb9f18b5d26 ("[PATCH] Linux-0.99.10 (June 7, 1993)") was a bugfix for this and the 2 didn't really mean O_RDWR, it meant FMODE_WRITE which was correct. The clue is that FMODE_{READ,WRITE} didn't exist yet and thus a raw "2" value was passed. Fast forward 5 years when around 2.2.4pre4 (February 16, 1999) this code was changed to: - dentry = open_namei(corefile,O_CREAT | 2 | O_TRUNC | O_NOFOLLOW, 0600); ... + file = filp_open(corefile,O_CREAT | 2 | O_TRUNC | O_NOFOLLOW, 0600); At this point the raw "2" should have become O_WRONLY again as filp_open() didn't take FMODE_{READ,WRITE} but O_{RDONLY,WRONLY,RDWR}. Another 17 years later, the code was changed again cementing the mistake and making it almost impossible to detect when commit 378c6520e7d2 ("fs/coredump: prevent fsuid=0 dumps into user-controlled directories") replaced the raw "2" with O_RDWR. And now, here we are with this patch that sent us on a quest to answer the big questions in life such as "Why are coredump files opened with O_RDWR?" and "Is it safe to just use O_WRONLY?". So with this commit we're reintroducing O_WRONLY again and bringing this code back to its original state when it was first introduced in commit ddc733f452e0 ("[PATCH] Linux-0.97 (August 1, 1992)") over 30 years ago. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru> Message-Id: <20230420120409.602576-1-vsementsov@yandex-team.ru> [brauner@kernel.org: completely rewritten commit message] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Christian Brauner <brauner@kernel.org>
2023-04-20 12:04:09 +00:00
int open_flags = O_CREAT | O_WRONLY | O_NOFOLLOW |
2016-03-22 21:25:36 +00:00
O_LARGEFILE | O_EXCL;
if (cprm.limit < binfmt->min_coredump)
goto fail_unlock;
fs: if a coredump already exists, unlink and recreate with O_EXCL It was possible for an attacking user to trick root (or another user) into writing his coredumps into an attacker-readable, pre-existing file using rename() or link(), causing the disclosure of secret data from the victim process' virtual memory. Depending on the configuration, it was also possible to trick root into overwriting system files with coredumps. Fix that issue by never writing coredumps into existing files. Requirements for the attack: - The attack only applies if the victim's process has a nonzero RLIMIT_CORE and is dumpable. - The attacker can trick the victim into coredumping into an attacker-writable directory D, either because the core_pattern is relative and the victim's cwd is attacker-writable or because an absolute core_pattern pointing to a world-writable directory is used. - The attacker has one of these: A: on a system with protected_hardlinks=0: execute access to a folder containing a victim-owned, attacker-readable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (In practice, there are lots of files that fulfill this condition, e.g. entries in Debian's /var/lib/dpkg/info/.) This does not apply to most Linux systems because most distros set protected_hardlinks=1. B: on a system with protected_hardlinks=1: execute access to a folder containing a victim-owned, attacker-readable and attacker-writable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (This seems to be uncommon.) C: on any system, independent of protected_hardlinks: write access to a non-sticky folder containing a victim-owned, attacker-readable file on the same partition as D (This seems to be uncommon.) The basic idea is that the attacker moves the victim-owned file to where he expects the victim process to dump its core. The victim process dumps its core into the existing file, and the attacker reads the coredump from it. If the attacker can't move the file because he does not have write access to the containing directory, he can instead link the file to a directory he controls, then wait for the original link to the file to be deleted (because the kernel checks that the link count of the corefile is 1). A less reliable variant that requires D to be non-sticky works with link() and does not require deletion of the original link: link() the file into D, but then unlink() it directly before the kernel performs the link count check. On systems with protected_hardlinks=0, this variant allows an attacker to not only gain information from coredumps, but also clobber existing, victim-writable files with coredumps. (This could theoretically lead to a privilege escalation.) Signed-off-by: Jann Horn <jann@thejh.net> Cc: Kees Cook <keescook@chromium.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-09 22:38:28 +00:00
if (need_suid_safe && cn.corename[0] != '/') {
printk(KERN_WARNING "Pid %d(%s) can only dump core "\
"to fully qualified path!\n",
task_tgid_vnr(current), current->comm);
printk(KERN_WARNING "Skipping core dump\n");
goto fail_unlock;
}
fs: if a coredump already exists, unlink and recreate with O_EXCL It was possible for an attacking user to trick root (or another user) into writing his coredumps into an attacker-readable, pre-existing file using rename() or link(), causing the disclosure of secret data from the victim process' virtual memory. Depending on the configuration, it was also possible to trick root into overwriting system files with coredumps. Fix that issue by never writing coredumps into existing files. Requirements for the attack: - The attack only applies if the victim's process has a nonzero RLIMIT_CORE and is dumpable. - The attacker can trick the victim into coredumping into an attacker-writable directory D, either because the core_pattern is relative and the victim's cwd is attacker-writable or because an absolute core_pattern pointing to a world-writable directory is used. - The attacker has one of these: A: on a system with protected_hardlinks=0: execute access to a folder containing a victim-owned, attacker-readable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (In practice, there are lots of files that fulfill this condition, e.g. entries in Debian's /var/lib/dpkg/info/.) This does not apply to most Linux systems because most distros set protected_hardlinks=1. B: on a system with protected_hardlinks=1: execute access to a folder containing a victim-owned, attacker-readable and attacker-writable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (This seems to be uncommon.) C: on any system, independent of protected_hardlinks: write access to a non-sticky folder containing a victim-owned, attacker-readable file on the same partition as D (This seems to be uncommon.) The basic idea is that the attacker moves the victim-owned file to where he expects the victim process to dump its core. The victim process dumps its core into the existing file, and the attacker reads the coredump from it. If the attacker can't move the file because he does not have write access to the containing directory, he can instead link the file to a directory he controls, then wait for the original link to the file to be deleted (because the kernel checks that the link count of the corefile is 1). A less reliable variant that requires D to be non-sticky works with link() and does not require deletion of the original link: link() the file into D, but then unlink() it directly before the kernel performs the link count check. On systems with protected_hardlinks=0, this variant allows an attacker to not only gain information from coredumps, but also clobber existing, victim-writable files with coredumps. (This could theoretically lead to a privilege escalation.) Signed-off-by: Jann Horn <jann@thejh.net> Cc: Kees Cook <keescook@chromium.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-09 22:38:28 +00:00
/*
* Unlink the file if it exists unless this is a SUID
* binary - in that case, we're running around with root
* privs and don't want to unlink another user's coredump.
*/
if (!need_suid_safe) {
/*
* If it doesn't exist, that's fine. If there's some
* other problem, we'll catch it at the filp_open().
*/
do_unlinkat(AT_FDCWD, getname_kernel(cn.corename));
fs: if a coredump already exists, unlink and recreate with O_EXCL It was possible for an attacking user to trick root (or another user) into writing his coredumps into an attacker-readable, pre-existing file using rename() or link(), causing the disclosure of secret data from the victim process' virtual memory. Depending on the configuration, it was also possible to trick root into overwriting system files with coredumps. Fix that issue by never writing coredumps into existing files. Requirements for the attack: - The attack only applies if the victim's process has a nonzero RLIMIT_CORE and is dumpable. - The attacker can trick the victim into coredumping into an attacker-writable directory D, either because the core_pattern is relative and the victim's cwd is attacker-writable or because an absolute core_pattern pointing to a world-writable directory is used. - The attacker has one of these: A: on a system with protected_hardlinks=0: execute access to a folder containing a victim-owned, attacker-readable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (In practice, there are lots of files that fulfill this condition, e.g. entries in Debian's /var/lib/dpkg/info/.) This does not apply to most Linux systems because most distros set protected_hardlinks=1. B: on a system with protected_hardlinks=1: execute access to a folder containing a victim-owned, attacker-readable and attacker-writable file on the same partition as D, and the victim-owned file will be deleted before the main part of the attack takes place. (This seems to be uncommon.) C: on any system, independent of protected_hardlinks: write access to a non-sticky folder containing a victim-owned, attacker-readable file on the same partition as D (This seems to be uncommon.) The basic idea is that the attacker moves the victim-owned file to where he expects the victim process to dump its core. The victim process dumps its core into the existing file, and the attacker reads the coredump from it. If the attacker can't move the file because he does not have write access to the containing directory, he can instead link the file to a directory he controls, then wait for the original link to the file to be deleted (because the kernel checks that the link count of the corefile is 1). A less reliable variant that requires D to be non-sticky works with link() and does not require deletion of the original link: link() the file into D, but then unlink() it directly before the kernel performs the link count check. On systems with protected_hardlinks=0, this variant allows an attacker to not only gain information from coredumps, but also clobber existing, victim-writable files with coredumps. (This could theoretically lead to a privilege escalation.) Signed-off-by: Jann Horn <jann@thejh.net> Cc: Kees Cook <keescook@chromium.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-09 22:38:28 +00:00
}
/*
* There is a race between unlinking and creating the
* file, but if that causes an EEXIST here, that's
* fine - another process raced with us while creating
* the corefile, and the other process won. To userspace,
* what matters is that at least one of the two processes
* writes its coredump successfully, not which one.
*/
2016-03-22 21:25:36 +00:00
if (need_suid_safe) {
/*
* Using user namespaces, normal user tasks can change
* their current->fs->root to point to arbitrary
* directories. Since the intention of the "only dump
* with a fully qualified path" rule is to control where
* coredumps may be placed using root privileges,
* current->fs->root must not be used. Instead, use the
* root directory of init_task.
*/
struct path root;
task_lock(&init_task);
get_fs_root(init_task.fs, &root);
task_unlock(&init_task);
cprm.file = file_open_root(&root, cn.corename,
open_flags, 0600);
2016-03-22 21:25:36 +00:00
path_put(&root);
} else {
cprm.file = filp_open(cn.corename, open_flags, 0600);
}
if (IS_ERR(cprm.file))
goto fail_unlock;
inode = file_inode(cprm.file);
if (inode->i_nlink > 1)
goto close_fail;
if (d_unhashed(cprm.file->f_path.dentry))
goto close_fail;
/*
* AK: actually i see no reason to not allow this for named
* pipes etc, but keep the previous behaviour for now.
*/
if (!S_ISREG(inode->i_mode))
goto close_fail;
/*
* Don't dump core if the filesystem changed owner or mode
* of the file during file creation. This is an issue when
* a process dumps core while its cwd is e.g. on a vfat
* filesystem.
*/
idmap = file_mnt_idmap(cprm.file);
if (!vfsuid_eq_kuid(i_uid_into_vfsuid(idmap, inode),
current_fsuid())) {
pr_info_ratelimited("Core dump to %s aborted: cannot preserve file owner\n",
cn.corename);
goto close_fail;
}
if ((inode->i_mode & 0677) != 0600) {
pr_info_ratelimited("Core dump to %s aborted: cannot preserve file permissions\n",
cn.corename);
goto close_fail;
}
if (!(cprm.file->f_mode & FMODE_CAN_WRITE))
goto close_fail;
if (do_truncate(idmap, cprm.file->f_path.dentry,
0, 0, cprm.file))
goto close_fail;
}
/* get us an unshared descriptor table; almost always a no-op */
coredump: Document coredump code exclusively used by cell spufs Oleg Nesterov recently asked[1] why is there an unshare_files in do_coredump. After digging through all of the callers of lookup_fd it turns out that it is arch/powerpc/platforms/cell/spufs/coredump.c:coredump_next_context that needs the unshare_files in do_coredump. Looking at the history[2] this code was also the only piece of coredump code that required the unshare_files when the unshare_files was added. Looking at that code it turns out that cell is also the only architecture that implements elf_coredump_extra_notes_size and elf_coredump_extra_notes_write. I looked at the gdb repo[3] support for cell has been removed[4] in binutils 2.34. Geoff Levand reports he is still getting questions on how to run modern kernels on the PS3, from people using 3rd party firmware so this code is not dead. According to Wikipedia the last PS3 shipped in Japan sometime in 2017. So it will probably be a little while before everyone's hardware dies. Add some comments briefly documenting the coredump code that exists only to support cell spufs to make it easier to understand the coredump code. Eventually the hardware will be dead, or their won't be userspace tools, or the coredump code will be refactored and it will be too difficult to update a dead architecture and these comments make it easy to tell where to pull to remove cell spufs support. [1] https://lkml.kernel.org/r/20201123175052.GA20279@redhat.com [2] 179e037fc137 ("do_coredump(): make sure that descriptor table isn't shared") [3] git://sourceware.org/git/binutils-gdb.git [4] abf516c6931a ("Remove Cell Broadband Engine debugging support"). Link: https://lkml.kernel.org/r/87h7pdnlzv.fsf_-_@x220.int.ebiederm.org Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
2020-11-25 21:51:32 +00:00
/* The cell spufs coredump code reads the file descriptor tables */
retval = unshare_files();
if (retval)
goto close_fail;
if (!dump_interrupted()) {
coredump: fix crash when umh is disabled Commit 64e90a8acb859 ("Introduce STATIC_USERMODEHELPER to mediate call_usermodehelper()") added the optiont to disable all call_usermodehelper() calls by setting STATIC_USERMODEHELPER_PATH to an empty string. When this is done, and crashdump is triggered, it will crash on null pointer dereference, since we make assumptions over what call_usermodehelper_exec() did. This has been reported by Sergey when one triggers a a coredump with the following configuration: ``` CONFIG_STATIC_USERMODEHELPER=y CONFIG_STATIC_USERMODEHELPER_PATH="" kernel.core_pattern = |/usr/lib/systemd/systemd-coredump %P %u %g %s %t %c %h %e ``` The way disabling the umh was designed was that call_usermodehelper_exec() would just return early, without an error. But coredump assumes certain variables are set up for us when this happens, and calls ile_start_write(cprm.file) with a NULL file. [ 2.819676] BUG: kernel NULL pointer dereference, address: 0000000000000020 [ 2.819859] #PF: supervisor read access in kernel mode [ 2.820035] #PF: error_code(0x0000) - not-present page [ 2.820188] PGD 0 P4D 0 [ 2.820305] Oops: 0000 [#1] SMP PTI [ 2.820436] CPU: 2 PID: 89 Comm: a Not tainted 5.7.0-rc1+ #7 [ 2.820680] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS ?-20190711_202441-buildvm-armv7-10.arm.fedoraproject.org-2.fc31 04/01/2014 [ 2.821150] RIP: 0010:do_coredump+0xd80/0x1060 [ 2.821385] Code: e8 95 11 ed ff 48 c7 c6 cc a7 b4 81 48 8d bd 28 ff ff ff 89 c2 e8 70 f1 ff ff 41 89 c2 85 c0 0f 84 72 f7 ff ff e9 b4 fe ff ff <48> 8b 57 20 0f b7 02 66 25 00 f0 66 3d 00 8 0 0f 84 9c 01 00 00 44 [ 2.822014] RSP: 0000:ffffc9000029bcb8 EFLAGS: 00010246 [ 2.822339] RAX: 0000000000000000 RBX: ffff88803f860000 RCX: 000000000000000a [ 2.822746] RDX: 0000000000000009 RSI: 0000000000000282 RDI: 0000000000000000 [ 2.823141] RBP: ffffc9000029bde8 R08: 0000000000000000 R09: ffffc9000029bc00 [ 2.823508] R10: 0000000000000001 R11: ffff88803dec90be R12: ffffffff81c39da0 [ 2.823902] R13: ffff88803de84400 R14: 0000000000000000 R15: 0000000000000000 [ 2.824285] FS: 00007fee08183540(0000) GS:ffff88803e480000(0000) knlGS:0000000000000000 [ 2.824767] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 2.825111] CR2: 0000000000000020 CR3: 000000003f856005 CR4: 0000000000060ea0 [ 2.825479] Call Trace: [ 2.825790] get_signal+0x11e/0x720 [ 2.826087] do_signal+0x1d/0x670 [ 2.826361] ? force_sig_info_to_task+0xc1/0xf0 [ 2.826691] ? force_sig_fault+0x3c/0x40 [ 2.826996] ? do_trap+0xc9/0x100 [ 2.827179] exit_to_usermode_loop+0x49/0x90 [ 2.827359] prepare_exit_to_usermode+0x77/0xb0 [ 2.827559] ? invalid_op+0xa/0x30 [ 2.827747] ret_from_intr+0x20/0x20 [ 2.827921] RIP: 0033:0x55e2c76d2129 [ 2.828107] Code: 2d ff ff ff e8 68 ff ff ff 5d c6 05 18 2f 00 00 01 c3 0f 1f 80 00 00 00 00 c3 0f 1f 80 00 00 00 00 e9 7b ff ff ff 55 48 89 e5 <0f> 0b b8 00 00 00 00 5d c3 66 2e 0f 1f 84 0 0 00 00 00 00 0f 1f 40 [ 2.828603] RSP: 002b:00007fffeba5e080 EFLAGS: 00010246 [ 2.828801] RAX: 000055e2c76d2125 RBX: 0000000000000000 RCX: 00007fee0817c718 [ 2.829034] RDX: 00007fffeba5e188 RSI: 00007fffeba5e178 RDI: 0000000000000001 [ 2.829257] RBP: 00007fffeba5e080 R08: 0000000000000000 R09: 00007fee08193c00 [ 2.829482] R10: 0000000000000009 R11: 0000000000000000 R12: 000055e2c76d2040 [ 2.829727] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000 [ 2.829964] CR2: 0000000000000020 [ 2.830149] ---[ end trace ceed83d8c68a1bf1 ]--- ``` Cc: <stable@vger.kernel.org> # v4.11+ Fixes: 64e90a8acb85 ("Introduce STATIC_USERMODEHELPER to mediate call_usermodehelper()") BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=199795 Reported-by: Tony Vroon <chainsaw@gentoo.org> Reported-by: Sergey Kvachonok <ravenexp@gmail.com> Tested-by: Sergei Trofimovich <slyfox@gentoo.org> Signed-off-by: Luis Chamberlain <mcgrof@kernel.org> Link: https://lore.kernel.org/r/20200416162859.26518-1-mcgrof@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2020-04-16 16:28:59 +00:00
/*
* umh disabled with CONFIG_STATIC_USERMODEHELPER_PATH="" would
* have this set to NULL.
*/
if (!cprm.file) {
pr_info("Core dump to |%s disabled\n", cn.corename);
goto close_fail;
}
if (!dump_vma_snapshot(&cprm))
goto close_fail;
file_start_write(cprm.file);
core_dumped = binfmt->core_dump(&cprm);
/*
* Ensures that file size is big enough to contain the current
* file postion. This prevents gdb from complaining about
* a truncated file if the last "write" to the file was
* dump_skip.
*/
if (cprm.to_skip) {
cprm.to_skip--;
dump_emit(&cprm, "", 1);
}
file_end_write(cprm.file);
free_vma_snapshot(&cprm);
}
if (ispipe && core_pipe_limit)
wait_for_dump_helpers(cprm.file);
close_fail:
if (cprm.file)
filp_close(cprm.file, NULL);
fail_dropcount:
if (ispipe)
atomic_dec(&core_dump_count);
fail_unlock:
coredump: split pipe command whitespace before expanding template Save the offsets of the start of each argument to avoid having to update pointers to each argument after every corename krealloc and to avoid having to duplicate the memory for the dump command. Executable names containing spaces were previously being expanded from %e or %E and then split in the middle of the filename. This is incorrect behaviour since an argument list can represent arguments with spaces. The splitting could lead to extra arguments being passed to the core dump handler that it might have interpreted as options or ignored completely. Core dump handlers that are not aware of this Linux kernel issue will be using %e or %E without considering that it may be split and so they will be vulnerable to processes with spaces in their names breaking their argument list. If their internals are otherwise well written, such as if they are written in shell but quote arguments, they will work better after this change than before. If they are not well written, then there is a slight chance of breakage depending on the details of the code but they will already be fairly broken by the split filenames. Core dump handlers that are aware of this Linux kernel issue will be placing %e or %E as the last item in their core_pattern and then aggregating all of the remaining arguments into one, separated by spaces. Alternatively they will be obtaining the filename via other methods. Both of these will be compatible with the new arrangement. A side effect from this change is that unknown template types (for example %z) result in an empty argument to the dump handler instead of the argument being dropped. This is a desired change as: It is easier for dump handlers to process empty arguments than dropped ones, especially if they are written in shell or don't pass each template item with a preceding command-line option in order to differentiate between individual template types. Most core_patterns in the wild do not use options so they can confuse different template types (especially numeric ones) if an earlier one gets dropped in old kernels. If the kernel introduces a new template type and a core_pattern uses it, the core dump handler might not expect that the argument can be dropped in old kernels. For example, this can result in security issues when %d is dropped in old kernels. This happened with the corekeeper package in Debian and resulted in the interface between corekeeper and Linux having to be rewritten to use command-line options to differentiate between template types. The core_pattern for most core dump handlers is written by the handler author who would generally not insert unknown template types so this change should be compatible with all the core dump handlers that exist. Link: http://lkml.kernel.org/r/20190528051142.24939-1-pabs3@bonedaddy.net Fixes: 74aadce98605 ("core_pattern: allow passing of arguments to user mode helper when core_pattern is a pipe") Signed-off-by: Paul Wise <pabs3@bonedaddy.net> Reported-by: Jakub Wilk <jwilk@jwilk.net> [https://bugs.debian.org/924398] Reported-by: Paul Wise <pabs3@bonedaddy.net> [https://lore.kernel.org/linux-fsdevel/c8b7ecb8508895bf4adb62a748e2ea2c71854597.camel@bonedaddy.net/] Suggested-by: Jakub Wilk <jwilk@jwilk.net> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-03 04:49:05 +00:00
kfree(argv);
kfree(cn.corename);
coredump: Limit coredumps to a single thread group Today when a signal is delivered with a handler of SIG_DFL whose default behavior is to generate a core dump not only that process but every process that shares the mm is killed. In the case of vfork this looks like a real world problem. Consider the following well defined sequence. if (vfork() == 0) { execve(...); _exit(EXIT_FAILURE); } If a signal that generates a core dump is received after vfork but before the execve changes the mm the process that called vfork will also be killed (as the mm is shared). Similarly if the execve fails after the point of no return the kernel delivers SIGSEGV which will kill both the exec'ing process and because the mm is shared the process that called vfork as well. As far as I can tell this behavior is a violation of people's reasonable expectations, POSIX, and is unnecessarily fragile when the system is low on memory. Solve this by making a userspace visible change to only kill a single process/thread group. This is possible because Jann Horn recently modified[1] the coredump code so that the mm can safely be modified while the coredump is happening. With LinuxThreads long gone I don't expect anyone to have a notice this behavior change in practice. To accomplish this move the core_state pointer from mm_struct to signal_struct, which allows different thread groups to coredump simultatenously. In zap_threads remove the work to kill anything except for the current thread group. v2: Remove core_state from the VM_BUG_ON_MM print to fix compile failure when CONFIG_DEBUG_VM is enabled. Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> [1] a07279c9a8cd ("binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot") Fixes: d89f3847def4 ("[PATCH] thread-aware coredumps, 2.5.43-C3") History-tree: git://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git Link: https://lkml.kernel.org/r/87y27mvnke.fsf@disp2133 Link: https://lkml.kernel.org/r/20211007144701.67592574@canb.auug.org.au Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
2021-09-22 16:24:02 +00:00
coredump_finish(core_dumped);
revert_creds(old_cred);
fail_creds:
put_cred(cred);
fail:
return;
}
/*
* Core dumping helper functions. These are the only things you should
* do on a core-file: use only these functions to write out all the
* necessary info.
*/
static int __dump_emit(struct coredump_params *cprm, const void *addr, int nr)
{
struct file *file = cprm->file;
loff_t pos = file->f_pos;
ssize_t n;
if (cprm->written + nr > cprm->limit)
return 0;
if (dump_interrupted())
return 0;
n = __kernel_write(file, addr, nr, &pos);
if (n != nr)
return 0;
file->f_pos = pos;
cprm->written += n;
cprm->pos += n;
return 1;
}
static int __dump_skip(struct coredump_params *cprm, size_t nr)
{
static char zeroes[PAGE_SIZE];
struct file *file = cprm->file;
if (file->f_mode & FMODE_LSEEK) {
coredump: introduce dump_interrupted() By discussion with Mandeep. Change dump_write(), dump_seek() and do_coredump() to check signal_pending() and abort if it is true. dump_seek() does this only before f_op->llseek(), otherwise it relies on dump_write(). We need this change to ensure that the coredump won't delay suspend, and to ensure it reacts to SIGKILL "quickly enough", a core dump can take a lot of time. In particular this can help oom-killer. We add the new trivial helper, dump_interrupted() to add the comments and to simplify the potential freezer changes. Perhaps it will have more callers. Ideally it should do try_to_freeze() but then we need the unpleasant changes in dump_write() and wait_for_dump_helpers(). It is not trivial to change dump_write() to restart if f_op->write() fails because of freezing(). We need to handle the short writes, we need to clear TIF_SIGPENDING (and we can't rely on recalc_sigpending() unless we change it to check PF_DUMPCORE). And if the buggy f_op->write() sets TIF_SIGPENDING we can not distinguish this case from the race with freeze_task() + __thaw_task(). So we simply accept the fact that the freezer can truncate a core-dump but at least you can reliably suspend. Hopefully we can tolerate this unlikely case and the necessary complications doesn't worth a trouble. But if we decide to make the coredumping freezable later we can do this on top of this change. Signed-off-by: Oleg Nesterov <oleg@redhat.com> Acked-by: Mandeep Singh Baines <msb@chromium.org> Cc: Neil Horman <nhorman@redhat.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Tejun Heo <tj@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-04-30 22:28:15 +00:00
if (dump_interrupted() ||
vfs_llseek(file, nr, SEEK_CUR) < 0)
return 0;
cprm->pos += nr;
return 1;
} else {
while (nr > PAGE_SIZE) {
if (!__dump_emit(cprm, zeroes, PAGE_SIZE))
return 0;
nr -= PAGE_SIZE;
}
return __dump_emit(cprm, zeroes, nr);
}
}
int dump_emit(struct coredump_params *cprm, const void *addr, int nr)
{
if (cprm->to_skip) {
if (!__dump_skip(cprm, cprm->to_skip))
return 0;
cprm->to_skip = 0;
}
return __dump_emit(cprm, addr, nr);
}
EXPORT_SYMBOL(dump_emit);
void dump_skip_to(struct coredump_params *cprm, unsigned long pos)
{
cprm->to_skip = pos - cprm->pos;
}
EXPORT_SYMBOL(dump_skip_to);
void dump_skip(struct coredump_params *cprm, size_t nr)
{
cprm->to_skip += nr;
}
EXPORT_SYMBOL(dump_skip);
#ifdef CONFIG_ELF_CORE
static int dump_emit_page(struct coredump_params *cprm, struct page *page)
{
struct bio_vec bvec;
struct iov_iter iter;
struct file *file = cprm->file;
loff_t pos;
ssize_t n;
iov_iter: get rid of 'copy_mc' flag This flag is only set by one single user: the magical core dumping code that looks up user pages one by one, and then writes them out using their kernel addresses (by using a BVEC_ITER). That actually ends up being a huge problem, because while we do use copy_mc_to_kernel() for this case and it is able to handle the possible machine checks involved, nothing else is really ready to handle the failures caused by the machine check. In particular, as reported by Tong Tiangen, we don't actually support fault_in_iov_iter_readable() on a machine check area. As a result, the usual logic for writing things to a file under a filesystem lock, which involves doing a copy with page faults disabled and then if that fails trying to fault pages in without holding the locks with fault_in_iov_iter_readable() does not work at all. We could decide to always just make the MC copy "succeed" (and filling the destination with zeroes), and that would then create a core dump file that just ignores any machine checks. But honestly, this single special case has been problematic before, and means that all the normal iov_iter code ends up slightly more complex and slower. See for example commit c9eec08bac96 ("iov_iter: Don't deal with iter->copy_mc in memcpy_from_iter_mc()") where David Howells re-organized the code just to avoid having to check the 'copy_mc' flags inside the inner iov_iter loops. So considering that we have exactly one user, and that one user is a non-critical special case that doesn't actually ever trigger in real life (Tong found this with manual error injection), the sane solution is to just decide that the onus on handling the machine check lines on that user instead. Ergo, do the copy_mc_to_kernel() in the core dump logic itself, copying the user data to a stable kernel page before writing it out. Fixes: f1982740f5e7 ("iov_iter: Convert iterate*() to inline funcs") Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Tong Tiangen <tongtiangen@huawei.com> Link: https://lore.kernel.org/r/20240305133336.3804360-1-tongtiangen@huawei.com Link: https://lore.kernel.org/all/4e80924d-9c85-f13a-722a-6a5d2b1c225a@huawei.com/ Tested-by: David Howells <dhowells@redhat.com> Reviewed-by: David Howells <dhowells@redhat.com> Reviewed-by: Jens Axboe <axboe@kernel.dk> Reported-by: Tong Tiangen <tongtiangen@huawei.com> Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-03-05 13:33:36 +00:00
if (!page)
return 0;
if (cprm->to_skip) {
if (!__dump_skip(cprm, cprm->to_skip))
return 0;
cprm->to_skip = 0;
}
if (cprm->written + PAGE_SIZE > cprm->limit)
return 0;
if (dump_interrupted())
return 0;
pos = file->f_pos;
bvec_set_page(&bvec, page, PAGE_SIZE, 0);
iov_iter_bvec(&iter, ITER_SOURCE, &bvec, 1, PAGE_SIZE);
n = __kernel_write_iter(cprm->file, &iter, &pos);
if (n != PAGE_SIZE)
return 0;
file->f_pos = pos;
cprm->written += PAGE_SIZE;
cprm->pos += PAGE_SIZE;
return 1;
}
iov_iter: get rid of 'copy_mc' flag This flag is only set by one single user: the magical core dumping code that looks up user pages one by one, and then writes them out using their kernel addresses (by using a BVEC_ITER). That actually ends up being a huge problem, because while we do use copy_mc_to_kernel() for this case and it is able to handle the possible machine checks involved, nothing else is really ready to handle the failures caused by the machine check. In particular, as reported by Tong Tiangen, we don't actually support fault_in_iov_iter_readable() on a machine check area. As a result, the usual logic for writing things to a file under a filesystem lock, which involves doing a copy with page faults disabled and then if that fails trying to fault pages in without holding the locks with fault_in_iov_iter_readable() does not work at all. We could decide to always just make the MC copy "succeed" (and filling the destination with zeroes), and that would then create a core dump file that just ignores any machine checks. But honestly, this single special case has been problematic before, and means that all the normal iov_iter code ends up slightly more complex and slower. See for example commit c9eec08bac96 ("iov_iter: Don't deal with iter->copy_mc in memcpy_from_iter_mc()") where David Howells re-organized the code just to avoid having to check the 'copy_mc' flags inside the inner iov_iter loops. So considering that we have exactly one user, and that one user is a non-critical special case that doesn't actually ever trigger in real life (Tong found this with manual error injection), the sane solution is to just decide that the onus on handling the machine check lines on that user instead. Ergo, do the copy_mc_to_kernel() in the core dump logic itself, copying the user data to a stable kernel page before writing it out. Fixes: f1982740f5e7 ("iov_iter: Convert iterate*() to inline funcs") Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Tong Tiangen <tongtiangen@huawei.com> Link: https://lore.kernel.org/r/20240305133336.3804360-1-tongtiangen@huawei.com Link: https://lore.kernel.org/all/4e80924d-9c85-f13a-722a-6a5d2b1c225a@huawei.com/ Tested-by: David Howells <dhowells@redhat.com> Reviewed-by: David Howells <dhowells@redhat.com> Reviewed-by: Jens Axboe <axboe@kernel.dk> Reported-by: Tong Tiangen <tongtiangen@huawei.com> Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-03-05 13:33:36 +00:00
/*
* If we might get machine checks from kernel accesses during the
* core dump, let's get those errors early rather than during the
* IO. This is not performance-critical enough to warrant having
* all the machine check logic in the iovec paths.
*/
#ifdef copy_mc_to_kernel
#define dump_page_alloc() alloc_page(GFP_KERNEL)
#define dump_page_free(x) __free_page(x)
static struct page *dump_page_copy(struct page *src, struct page *dst)
{
void *buf = kmap_local_page(src);
size_t left = copy_mc_to_kernel(page_address(dst), buf, PAGE_SIZE);
kunmap_local(buf);
return left ? NULL : dst;
}
#else
/* We just want to return non-NULL; it's never used. */
#define dump_page_alloc() ERR_PTR(-EINVAL)
#define dump_page_free(x) ((void)(x))
static inline struct page *dump_page_copy(struct page *src, struct page *dst)
{
return src;
}
#endif
int dump_user_range(struct coredump_params *cprm, unsigned long start,
unsigned long len)
{
unsigned long addr;
iov_iter: get rid of 'copy_mc' flag This flag is only set by one single user: the magical core dumping code that looks up user pages one by one, and then writes them out using their kernel addresses (by using a BVEC_ITER). That actually ends up being a huge problem, because while we do use copy_mc_to_kernel() for this case and it is able to handle the possible machine checks involved, nothing else is really ready to handle the failures caused by the machine check. In particular, as reported by Tong Tiangen, we don't actually support fault_in_iov_iter_readable() on a machine check area. As a result, the usual logic for writing things to a file under a filesystem lock, which involves doing a copy with page faults disabled and then if that fails trying to fault pages in without holding the locks with fault_in_iov_iter_readable() does not work at all. We could decide to always just make the MC copy "succeed" (and filling the destination with zeroes), and that would then create a core dump file that just ignores any machine checks. But honestly, this single special case has been problematic before, and means that all the normal iov_iter code ends up slightly more complex and slower. See for example commit c9eec08bac96 ("iov_iter: Don't deal with iter->copy_mc in memcpy_from_iter_mc()") where David Howells re-organized the code just to avoid having to check the 'copy_mc' flags inside the inner iov_iter loops. So considering that we have exactly one user, and that one user is a non-critical special case that doesn't actually ever trigger in real life (Tong found this with manual error injection), the sane solution is to just decide that the onus on handling the machine check lines on that user instead. Ergo, do the copy_mc_to_kernel() in the core dump logic itself, copying the user data to a stable kernel page before writing it out. Fixes: f1982740f5e7 ("iov_iter: Convert iterate*() to inline funcs") Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Tong Tiangen <tongtiangen@huawei.com> Link: https://lore.kernel.org/r/20240305133336.3804360-1-tongtiangen@huawei.com Link: https://lore.kernel.org/all/4e80924d-9c85-f13a-722a-6a5d2b1c225a@huawei.com/ Tested-by: David Howells <dhowells@redhat.com> Reviewed-by: David Howells <dhowells@redhat.com> Reviewed-by: Jens Axboe <axboe@kernel.dk> Reported-by: Tong Tiangen <tongtiangen@huawei.com> Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-03-05 13:33:36 +00:00
struct page *dump_page;
dump_page = dump_page_alloc();
if (!dump_page)
return 0;
for (addr = start; addr < start + len; addr += PAGE_SIZE) {
struct page *page;
/*
* To avoid having to allocate page tables for virtual address
* ranges that have never been used yet, and also to make it
* easy to generate sparse core files, use a helper that returns
* NULL when encountering an empty page table entry that would
* otherwise have been filled with the zero page.
*/
page = get_dump_page(addr);
if (page) {
iov_iter: get rid of 'copy_mc' flag This flag is only set by one single user: the magical core dumping code that looks up user pages one by one, and then writes them out using their kernel addresses (by using a BVEC_ITER). That actually ends up being a huge problem, because while we do use copy_mc_to_kernel() for this case and it is able to handle the possible machine checks involved, nothing else is really ready to handle the failures caused by the machine check. In particular, as reported by Tong Tiangen, we don't actually support fault_in_iov_iter_readable() on a machine check area. As a result, the usual logic for writing things to a file under a filesystem lock, which involves doing a copy with page faults disabled and then if that fails trying to fault pages in without holding the locks with fault_in_iov_iter_readable() does not work at all. We could decide to always just make the MC copy "succeed" (and filling the destination with zeroes), and that would then create a core dump file that just ignores any machine checks. But honestly, this single special case has been problematic before, and means that all the normal iov_iter code ends up slightly more complex and slower. See for example commit c9eec08bac96 ("iov_iter: Don't deal with iter->copy_mc in memcpy_from_iter_mc()") where David Howells re-organized the code just to avoid having to check the 'copy_mc' flags inside the inner iov_iter loops. So considering that we have exactly one user, and that one user is a non-critical special case that doesn't actually ever trigger in real life (Tong found this with manual error injection), the sane solution is to just decide that the onus on handling the machine check lines on that user instead. Ergo, do the copy_mc_to_kernel() in the core dump logic itself, copying the user data to a stable kernel page before writing it out. Fixes: f1982740f5e7 ("iov_iter: Convert iterate*() to inline funcs") Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Tong Tiangen <tongtiangen@huawei.com> Link: https://lore.kernel.org/r/20240305133336.3804360-1-tongtiangen@huawei.com Link: https://lore.kernel.org/all/4e80924d-9c85-f13a-722a-6a5d2b1c225a@huawei.com/ Tested-by: David Howells <dhowells@redhat.com> Reviewed-by: David Howells <dhowells@redhat.com> Reviewed-by: Jens Axboe <axboe@kernel.dk> Reported-by: Tong Tiangen <tongtiangen@huawei.com> Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-03-05 13:33:36 +00:00
int stop = !dump_emit_page(cprm, dump_page_copy(page, dump_page));
put_page(page);
iov_iter: get rid of 'copy_mc' flag This flag is only set by one single user: the magical core dumping code that looks up user pages one by one, and then writes them out using their kernel addresses (by using a BVEC_ITER). That actually ends up being a huge problem, because while we do use copy_mc_to_kernel() for this case and it is able to handle the possible machine checks involved, nothing else is really ready to handle the failures caused by the machine check. In particular, as reported by Tong Tiangen, we don't actually support fault_in_iov_iter_readable() on a machine check area. As a result, the usual logic for writing things to a file under a filesystem lock, which involves doing a copy with page faults disabled and then if that fails trying to fault pages in without holding the locks with fault_in_iov_iter_readable() does not work at all. We could decide to always just make the MC copy "succeed" (and filling the destination with zeroes), and that would then create a core dump file that just ignores any machine checks. But honestly, this single special case has been problematic before, and means that all the normal iov_iter code ends up slightly more complex and slower. See for example commit c9eec08bac96 ("iov_iter: Don't deal with iter->copy_mc in memcpy_from_iter_mc()") where David Howells re-organized the code just to avoid having to check the 'copy_mc' flags inside the inner iov_iter loops. So considering that we have exactly one user, and that one user is a non-critical special case that doesn't actually ever trigger in real life (Tong found this with manual error injection), the sane solution is to just decide that the onus on handling the machine check lines on that user instead. Ergo, do the copy_mc_to_kernel() in the core dump logic itself, copying the user data to a stable kernel page before writing it out. Fixes: f1982740f5e7 ("iov_iter: Convert iterate*() to inline funcs") Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Tong Tiangen <tongtiangen@huawei.com> Link: https://lore.kernel.org/r/20240305133336.3804360-1-tongtiangen@huawei.com Link: https://lore.kernel.org/all/4e80924d-9c85-f13a-722a-6a5d2b1c225a@huawei.com/ Tested-by: David Howells <dhowells@redhat.com> Reviewed-by: David Howells <dhowells@redhat.com> Reviewed-by: Jens Axboe <axboe@kernel.dk> Reported-by: Tong Tiangen <tongtiangen@huawei.com> Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-03-05 13:33:36 +00:00
if (stop) {
dump_page_free(dump_page);
return 0;
iov_iter: get rid of 'copy_mc' flag This flag is only set by one single user: the magical core dumping code that looks up user pages one by one, and then writes them out using their kernel addresses (by using a BVEC_ITER). That actually ends up being a huge problem, because while we do use copy_mc_to_kernel() for this case and it is able to handle the possible machine checks involved, nothing else is really ready to handle the failures caused by the machine check. In particular, as reported by Tong Tiangen, we don't actually support fault_in_iov_iter_readable() on a machine check area. As a result, the usual logic for writing things to a file under a filesystem lock, which involves doing a copy with page faults disabled and then if that fails trying to fault pages in without holding the locks with fault_in_iov_iter_readable() does not work at all. We could decide to always just make the MC copy "succeed" (and filling the destination with zeroes), and that would then create a core dump file that just ignores any machine checks. But honestly, this single special case has been problematic before, and means that all the normal iov_iter code ends up slightly more complex and slower. See for example commit c9eec08bac96 ("iov_iter: Don't deal with iter->copy_mc in memcpy_from_iter_mc()") where David Howells re-organized the code just to avoid having to check the 'copy_mc' flags inside the inner iov_iter loops. So considering that we have exactly one user, and that one user is a non-critical special case that doesn't actually ever trigger in real life (Tong found this with manual error injection), the sane solution is to just decide that the onus on handling the machine check lines on that user instead. Ergo, do the copy_mc_to_kernel() in the core dump logic itself, copying the user data to a stable kernel page before writing it out. Fixes: f1982740f5e7 ("iov_iter: Convert iterate*() to inline funcs") Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Tong Tiangen <tongtiangen@huawei.com> Link: https://lore.kernel.org/r/20240305133336.3804360-1-tongtiangen@huawei.com Link: https://lore.kernel.org/all/4e80924d-9c85-f13a-722a-6a5d2b1c225a@huawei.com/ Tested-by: David Howells <dhowells@redhat.com> Reviewed-by: David Howells <dhowells@redhat.com> Reviewed-by: Jens Axboe <axboe@kernel.dk> Reported-by: Tong Tiangen <tongtiangen@huawei.com> Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-03-05 13:33:36 +00:00
}
} else {
dump_skip(cprm, PAGE_SIZE);
}
}
iov_iter: get rid of 'copy_mc' flag This flag is only set by one single user: the magical core dumping code that looks up user pages one by one, and then writes them out using their kernel addresses (by using a BVEC_ITER). That actually ends up being a huge problem, because while we do use copy_mc_to_kernel() for this case and it is able to handle the possible machine checks involved, nothing else is really ready to handle the failures caused by the machine check. In particular, as reported by Tong Tiangen, we don't actually support fault_in_iov_iter_readable() on a machine check area. As a result, the usual logic for writing things to a file under a filesystem lock, which involves doing a copy with page faults disabled and then if that fails trying to fault pages in without holding the locks with fault_in_iov_iter_readable() does not work at all. We could decide to always just make the MC copy "succeed" (and filling the destination with zeroes), and that would then create a core dump file that just ignores any machine checks. But honestly, this single special case has been problematic before, and means that all the normal iov_iter code ends up slightly more complex and slower. See for example commit c9eec08bac96 ("iov_iter: Don't deal with iter->copy_mc in memcpy_from_iter_mc()") where David Howells re-organized the code just to avoid having to check the 'copy_mc' flags inside the inner iov_iter loops. So considering that we have exactly one user, and that one user is a non-critical special case that doesn't actually ever trigger in real life (Tong found this with manual error injection), the sane solution is to just decide that the onus on handling the machine check lines on that user instead. Ergo, do the copy_mc_to_kernel() in the core dump logic itself, copying the user data to a stable kernel page before writing it out. Fixes: f1982740f5e7 ("iov_iter: Convert iterate*() to inline funcs") Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Tong Tiangen <tongtiangen@huawei.com> Link: https://lore.kernel.org/r/20240305133336.3804360-1-tongtiangen@huawei.com Link: https://lore.kernel.org/all/4e80924d-9c85-f13a-722a-6a5d2b1c225a@huawei.com/ Tested-by: David Howells <dhowells@redhat.com> Reviewed-by: David Howells <dhowells@redhat.com> Reviewed-by: Jens Axboe <axboe@kernel.dk> Reported-by: Tong Tiangen <tongtiangen@huawei.com> Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-03-05 13:33:36 +00:00
dump_page_free(dump_page);
return 1;
}
#endif
int dump_align(struct coredump_params *cprm, int align)
{
unsigned mod = (cprm->pos + cprm->to_skip) & (align - 1);
if (align & (align - 1))
return 0;
if (mod)
cprm->to_skip += align - mod;
return 1;
}
EXPORT_SYMBOL(dump_align);
#ifdef CONFIG_SYSCTL
void validate_coredump_safety(void)
{
if (suid_dumpable == SUID_DUMP_ROOT &&
core_pattern[0] != '/' && core_pattern[0] != '|') {
pr_warn(
"Unsafe core_pattern used with fs.suid_dumpable=2.\n"
"Pipe handler or fully qualified core dump path required.\n"
"Set kernel.core_pattern before fs.suid_dumpable.\n"
);
}
}
sysctl: treewide: constify the ctl_table argument of proc_handlers const qualify the struct ctl_table argument in the proc_handler function signatures. This is a prerequisite to moving the static ctl_table structs into .rodata data which will ensure that proc_handler function pointers cannot be modified. This patch has been generated by the following coccinelle script: ``` virtual patch @r1@ identifier ctl, write, buffer, lenp, ppos; identifier func !~ "appldata_(timer|interval)_handler|sched_(rt|rr)_handler|rds_tcp_skbuf_handler|proc_sctp_do_(hmac_alg|rto_min|rto_max|udp_port|alpha_beta|auth|probe_interval)"; @@ int func( - struct ctl_table *ctl + const struct ctl_table *ctl ,int write, void *buffer, size_t *lenp, loff_t *ppos); @r2@ identifier func, ctl, write, buffer, lenp, ppos; @@ int func( - struct ctl_table *ctl + const struct ctl_table *ctl ,int write, void *buffer, size_t *lenp, loff_t *ppos) { ... } @r3@ identifier func; @@ int func( - struct ctl_table * + const struct ctl_table * ,int , void *, size_t *, loff_t *); @r4@ identifier func, ctl; @@ int func( - struct ctl_table *ctl + const struct ctl_table *ctl ,int , void *, size_t *, loff_t *); @r5@ identifier func, write, buffer, lenp, ppos; @@ int func( - struct ctl_table * + const struct ctl_table * ,int write, void *buffer, size_t *lenp, loff_t *ppos); ``` * Code formatting was adjusted in xfs_sysctl.c to comply with code conventions. The xfs_stats_clear_proc_handler, xfs_panic_mask_proc_handler and xfs_deprecated_dointvec_minmax where adjusted. * The ctl_table argument in proc_watchdog_common was const qualified. This is called from a proc_handler itself and is calling back into another proc_handler, making it necessary to change it as part of the proc_handler migration. Co-developed-by: Thomas Weißschuh <linux@weissschuh.net> Signed-off-by: Thomas Weißschuh <linux@weissschuh.net> Co-developed-by: Joel Granados <j.granados@samsung.com> Signed-off-by: Joel Granados <j.granados@samsung.com>
2024-07-24 18:59:29 +00:00
static int proc_dostring_coredump(const struct ctl_table *table, int write,
void *buffer, size_t *lenp, loff_t *ppos)
{
int error = proc_dostring(table, write, buffer, lenp, ppos);
if (!error)
validate_coredump_safety();
return error;
}
static const unsigned int core_file_note_size_min = CORE_FILE_NOTE_SIZE_DEFAULT;
static const unsigned int core_file_note_size_max = CORE_FILE_NOTE_SIZE_MAX;
static struct ctl_table coredump_sysctls[] = {
{
.procname = "core_uses_pid",
.data = &core_uses_pid,
.maxlen = sizeof(int),
.mode = 0644,
.proc_handler = proc_dointvec,
},
{
.procname = "core_pattern",
.data = core_pattern,
.maxlen = CORENAME_MAX_SIZE,
.mode = 0644,
.proc_handler = proc_dostring_coredump,
},
{
.procname = "core_pipe_limit",
.data = &core_pipe_limit,
.maxlen = sizeof(unsigned int),
.mode = 0644,
.proc_handler = proc_dointvec,
},
{
.procname = "core_file_note_size_limit",
.data = &core_file_note_size_limit,
.maxlen = sizeof(unsigned int),
.mode = 0644,
.proc_handler = proc_douintvec_minmax,
.extra1 = (unsigned int *)&core_file_note_size_min,
.extra2 = (unsigned int *)&core_file_note_size_max,
},
};
static int __init init_fs_coredump_sysctls(void)
{
register_sysctl_init("kernel", coredump_sysctls);
return 0;
}
fs_initcall(init_fs_coredump_sysctls);
#endif /* CONFIG_SYSCTL */
/*
* The purpose of always_dump_vma() is to make sure that special kernel mappings
* that are useful for post-mortem analysis are included in every core dump.
* In that way we ensure that the core dump is fully interpretable later
* without matching up the same kernel and hardware config to see what PC values
* meant. These special mappings include - vDSO, vsyscall, and other
* architecture specific mappings
*/
static bool always_dump_vma(struct vm_area_struct *vma)
{
/* Any vsyscall mappings? */
if (vma == get_gate_vma(vma->vm_mm))
return true;
/*
* Assume that all vmas with a .name op should always be dumped.
* If this changes, a new vm_ops field can easily be added.
*/
if (vma->vm_ops && vma->vm_ops->name && vma->vm_ops->name(vma))
return true;
/*
* arch_vma_name() returns non-NULL for special architecture mappings,
* such as vDSO sections.
*/
if (arch_vma_name(vma))
return true;
return false;
}
#define DUMP_SIZE_MAYBE_ELFHDR_PLACEHOLDER 1
/*
* Decide how much of @vma's contents should be included in a core dump.
*/
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
static unsigned long vma_dump_size(struct vm_area_struct *vma,
unsigned long mm_flags)
{
#define FILTER(type) (mm_flags & (1UL << MMF_DUMP_##type))
/* always dump the vdso and vsyscall sections */
if (always_dump_vma(vma))
goto whole;
if (vma->vm_flags & VM_DONTDUMP)
return 0;
/* support for DAX */
if (vma_is_dax(vma)) {
if ((vma->vm_flags & VM_SHARED) && FILTER(DAX_SHARED))
goto whole;
if (!(vma->vm_flags & VM_SHARED) && FILTER(DAX_PRIVATE))
goto whole;
return 0;
}
/* Hugetlb memory check */
if (is_vm_hugetlb_page(vma)) {
if ((vma->vm_flags & VM_SHARED) && FILTER(HUGETLB_SHARED))
goto whole;
if (!(vma->vm_flags & VM_SHARED) && FILTER(HUGETLB_PRIVATE))
goto whole;
return 0;
}
/* Do not dump I/O mapped devices or special mappings */
if (vma->vm_flags & VM_IO)
return 0;
/* By default, dump shared memory if mapped from an anonymous file. */
if (vma->vm_flags & VM_SHARED) {
if (file_inode(vma->vm_file)->i_nlink == 0 ?
FILTER(ANON_SHARED) : FILTER(MAPPED_SHARED))
goto whole;
return 0;
}
/* Dump segments that have been written to. */
if ((!IS_ENABLED(CONFIG_MMU) || vma->anon_vma) && FILTER(ANON_PRIVATE))
goto whole;
if (vma->vm_file == NULL)
return 0;
if (FILTER(MAPPED_PRIVATE))
goto whole;
/*
* If this is the beginning of an executable file mapping,
* dump the first page to aid in determining what was mapped here.
*/
if (FILTER(ELF_HEADERS) &&
vma->vm_pgoff == 0 && (vma->vm_flags & VM_READ)) {
if ((READ_ONCE(file_inode(vma->vm_file)->i_mode) & 0111) != 0)
return PAGE_SIZE;
/*
* ELF libraries aren't always executable.
* We'll want to check whether the mapping starts with the ELF
* magic, but not now - we're holding the mmap lock,
* so copy_from_user() doesn't work here.
* Use a placeholder instead, and fix it up later in
* dump_vma_snapshot().
*/
return DUMP_SIZE_MAYBE_ELFHDR_PLACEHOLDER;
}
#undef FILTER
return 0;
whole:
return vma->vm_end - vma->vm_start;
}
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
/*
* Helper function for iterating across a vma list. It ensures that the caller
* will visit `gate_vma' prior to terminating the search.
*/
static struct vm_area_struct *coredump_next_vma(struct vma_iterator *vmi,
struct vm_area_struct *vma,
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
struct vm_area_struct *gate_vma)
{
if (gate_vma && (vma == gate_vma))
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
return NULL;
vma = vma_next(vmi);
if (vma)
return vma;
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
return gate_vma;
}
static void free_vma_snapshot(struct coredump_params *cprm)
{
if (cprm->vma_meta) {
int i;
for (i = 0; i < cprm->vma_count; i++) {
struct file *file = cprm->vma_meta[i].file;
if (file)
fput(file);
}
kvfree(cprm->vma_meta);
cprm->vma_meta = NULL;
}
}
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
/*
* Under the mmap_lock, take a snapshot of relevant information about the task's
* VMAs.
*/
static bool dump_vma_snapshot(struct coredump_params *cprm)
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
{
struct vm_area_struct *gate_vma, *vma = NULL;
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
struct mm_struct *mm = current->mm;
VMA_ITERATOR(vmi, mm, 0);
int i = 0;
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
/*
* Once the stack expansion code is fixed to not change VMA bounds
* under mmap_lock in read mode, this can be changed to take the
* mmap_lock in read mode.
*/
if (mmap_write_lock_killable(mm))
return false;
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
cprm->vma_data_size = 0;
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
gate_vma = get_gate_vma(mm);
cprm->vma_count = mm->map_count + (gate_vma ? 1 : 0);
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
cprm->vma_meta = kvmalloc_array(cprm->vma_count, sizeof(*cprm->vma_meta), GFP_KERNEL);
if (!cprm->vma_meta) {
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
mmap_write_unlock(mm);
return false;
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
}
while ((vma = coredump_next_vma(&vmi, vma, gate_vma)) != NULL) {
struct core_vma_metadata *m = cprm->vma_meta + i;
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
m->start = vma->vm_start;
m->end = vma->vm_end;
m->flags = vma->vm_flags;
m->dump_size = vma_dump_size(vma, cprm->mm_flags);
m->pgoff = vma->vm_pgoff;
m->file = vma->vm_file;
if (m->file)
get_file(m->file);
i++;
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
}
mmap_write_unlock(mm);
for (i = 0; i < cprm->vma_count; i++) {
struct core_vma_metadata *m = cprm->vma_meta + i;
if (m->dump_size == DUMP_SIZE_MAYBE_ELFHDR_PLACEHOLDER) {
char elfmag[SELFMAG];
if (copy_from_user(elfmag, (void __user *)m->start, SELFMAG) ||
memcmp(elfmag, ELFMAG, SELFMAG) != 0) {
m->dump_size = 0;
} else {
m->dump_size = PAGE_SIZE;
}
}
cprm->vma_data_size += m->dump_size;
}
return true;
binfmt_elf, binfmt_elf_fdpic: use a VMA list snapshot In both binfmt_elf and binfmt_elf_fdpic, use a new helper dump_vma_snapshot() to take a snapshot of the VMA list (including the gate VMA, if we have one) while protected by the mmap_lock, and then use that snapshot instead of walking the VMA list without locking. An alternative approach would be to keep the mmap_lock held across the entire core dumping operation; however, keeping the mmap_lock locked while we may be blocked for an unbounded amount of time (e.g. because we're dumping to a FUSE filesystem or so) isn't really optimal; the mmap_lock blocks things like the ->release handler of userfaultfd, and we don't really want critical system daemons to grind to a halt just because someone "gifted" them SCM_RIGHTS to an eternally-locked userfaultfd, or something like that. Since both the normal ELF code and the FDPIC ELF code need this functionality (and if any other binfmt wants to add coredump support in the future, they'd probably need it, too), implement this with a common helper in fs/coredump.c. A downside of this approach is that we now need a bigger amount of kernel memory per userspace VMA in the normal ELF case, and that we need O(n) kernel memory in the FDPIC ELF case at all; but 40 bytes per VMA shouldn't be terribly bad. There currently is a data race between stack expansion and anything that reads ->vm_start or ->vm_end under the mmap_lock held in read mode; to mitigate that for core dumping, take the mmap_lock in write mode when taking a snapshot of the VMA hierarchy. (If we only took the mmap_lock in read mode, we could end up with a corrupted core dump if someone does get_user_pages_remote() concurrently. Not really a major problem, but taking the mmap_lock either way works here, so we might as well avoid the issue.) (This doesn't do anything about the existing data races with stack expansion in other mm code.) Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: "Eric W . Biederman" <ebiederm@xmission.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Hugh Dickins <hughd@google.com> Link: http://lkml.kernel.org/r/20200827114932.3572699-6-jannh@google.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-10-16 03:12:54 +00:00
}