When we attach any cgroup hook, the rest (even if unused/unattached) start
to contribute small overhead. In particular, the one we want to avoid is
__cgroup_bpf_run_filter_skb which does two redirections to get to
the cgroup and pushes/pulls skb.
Let's split cgroup_bpf_enabled to be per-attach to make sure
only used attach types trigger.
I've dropped some existing high-level cgroup_bpf_enabled in some
places because BPF_PROG_CGROUP_XXX_RUN macros usually have another
cgroup_bpf_enabled check.
I also had to copy-paste BPF_CGROUP_RUN_SA_PROG_LOCK for
GETPEERNAME/GETSOCKNAME because type for cgroup_bpf_enabled[type]
has to be constant and known at compile time.
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20210115163501.805133-4-sdf@google.com
When we attach a bpf program to cgroup/getsockopt any other getsockopt()
syscall starts incurring kzalloc/kfree cost.
Let add a small buffer on the stack and use it for small (majority)
{s,g}etsockopt values. The buffer is small enough to fit into
the cache line and cover the majority of simple options (most
of them are 4 byte ints).
It seems natural to do the same for setsockopt, but it's a bit more
involved when the BPF program modifies the data (where we have to
kmalloc). The assumption is that for the majority of setsockopt
calls (which are doing pure BPF options or apply policy) this
will bring some benefit as well.
Without this patch (we remove about 1% __kmalloc):
3.38% 0.07% tcp_mmap [kernel.kallsyms] [k] __cgroup_bpf_run_filter_getsockopt
|
--3.30%--__cgroup_bpf_run_filter_getsockopt
|
--0.81%--__kmalloc
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20210115163501.805133-3-sdf@google.com
Add custom implementation of getsockopt hook for TCP_ZEROCOPY_RECEIVE.
We skip generic hooks for TCP_ZEROCOPY_RECEIVE and have a custom
call in do_tcp_getsockopt using the on-stack data. This removes
3% overhead for locking/unlocking the socket.
Without this patch:
3.38% 0.07% tcp_mmap [kernel.kallsyms] [k] __cgroup_bpf_run_filter_getsockopt
|
--3.30%--__cgroup_bpf_run_filter_getsockopt
|
--0.81%--__kmalloc
With the patch applied:
0.52% 0.12% tcp_mmap [kernel.kallsyms] [k] __cgroup_bpf_run_filter_getsockopt_kern
Note, exporting uapi/tcp.h requires removing netinet/tcp.h
from test_progs.h because those headers have confliciting
definitions.
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20210115163501.805133-2-sdf@google.com
llvm patch https://reviews.llvm.org/D84002 permitted
to emit empty rodata datasec if the elf .rodata section
contains read-only data from local variables. These
local variables will be not emitted as BTF_KIND_VARs
since llvm converted these local variables as
static variables with private linkage without debuginfo
types. Such an empty rodata datasec will make
skeleton code generation easy since for skeleton
a rodata struct will be generated if there is a
.rodata elf section. The existence of a rodata
btf datasec is also consistent with the existence
of a rodata map created by libbpf.
The btf with such an empty rodata datasec will fail
in the kernel though as kernel will reject a datasec
with zero vlen and zero size. For example, for the below code,
int sys_enter(void *ctx)
{
int fmt[6] = {1, 2, 3, 4, 5, 6};
int dst[6];
bpf_probe_read(dst, sizeof(dst), fmt);
return 0;
}
We got the below btf (bpftool btf dump ./test.o):
[1] PTR '(anon)' type_id=0
[2] FUNC_PROTO '(anon)' ret_type_id=3 vlen=1
'ctx' type_id=1
[3] INT 'int' size=4 bits_offset=0 nr_bits=32 encoding=SIGNED
[4] FUNC 'sys_enter' type_id=2 linkage=global
[5] INT 'char' size=1 bits_offset=0 nr_bits=8 encoding=SIGNED
[6] ARRAY '(anon)' type_id=5 index_type_id=7 nr_elems=4
[7] INT '__ARRAY_SIZE_TYPE__' size=4 bits_offset=0 nr_bits=32 encoding=(none)
[8] VAR '_license' type_id=6, linkage=global-alloc
[9] DATASEC '.rodata' size=0 vlen=0
[10] DATASEC 'license' size=0 vlen=1
type_id=8 offset=0 size=4
When loading the ./test.o to the kernel with bpftool,
we see the following error:
libbpf: Error loading BTF: Invalid argument(22)
libbpf: magic: 0xeb9f
...
[6] ARRAY (anon) type_id=5 index_type_id=7 nr_elems=4
[7] INT __ARRAY_SIZE_TYPE__ size=4 bits_offset=0 nr_bits=32 encoding=(none)
[8] VAR _license type_id=6 linkage=1
[9] DATASEC .rodata size=24 vlen=0 vlen == 0
libbpf: Error loading .BTF into kernel: -22. BTF is optional, ignoring.
Basically, libbpf changed .rodata datasec size to 24 since elf .rodata
section size is 24. The kernel then rejected the BTF since vlen = 0.
Note that the above kernel verifier failure can be worked around with
changing local variable "fmt" to a static or global, optionally const, variable.
This patch permits a datasec with vlen = 0 in kernel.
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210119153519.3901963-1-yhs@fb.com
Conflicts:
drivers/net/can/dev.c
commit 03f16c5075 ("can: dev: can_restart: fix use after free bug")
commit 3e77f70e73 ("can: dev: move driver related infrastructure into separate subdir")
Code move.
drivers/net/dsa/b53/b53_common.c
commit 8e4052c32d ("net: dsa: b53: fix an off by one in checking "vlan->vid"")
commit b7a9e0da2d ("net: switchdev: remove vid_begin -> vid_end range from VLAN objects")
Field rename.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull networking fixes from Jakub Kicinski:
"Networking fixes for 5.11-rc5, including fixes from bpf, wireless, and
can trees.
Current release - regressions:
- nfc: nci: fix the wrong NCI_CORE_INIT parameters
Current release - new code bugs:
- bpf: allow empty module BTFs
Previous releases - regressions:
- bpf: fix signed_{sub,add32}_overflows type handling
- tcp: do not mess with cloned skbs in tcp_add_backlog()
- bpf: prevent double bpf_prog_put call from bpf_tracing_prog_attach
- bpf: don't leak memory in bpf getsockopt when optlen == 0
- tcp: fix potential use-after-free due to double kfree()
- mac80211: fix encryption issues with WEP
- devlink: use right genl user_ptr when handling port param get/set
- ipv6: set multicast flag on the multicast route
- tcp: fix TCP_USER_TIMEOUT with zero window
Previous releases - always broken:
- bpf: local storage helpers should check nullness of owner ptr passed
- mac80211: fix incorrect strlen of .write in debugfs
- cls_flower: call nla_ok() before nla_next()
- skbuff: back tiny skbs with kmalloc() in __netdev_alloc_skb() too"
* tag 'net-5.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (52 commits)
net: systemport: free dev before on error path
net: usb: cdc_ncm: don't spew notifications
net: mscc: ocelot: Fix multicast to the CPU port
tcp: Fix potential use-after-free due to double kfree()
bpf: Fix signed_{sub,add32}_overflows type handling
can: peak_usb: fix use after free bugs
can: vxcan: vxcan_xmit: fix use after free bug
can: dev: can_restart: fix use after free bug
tcp: fix TCP socket rehash stats mis-accounting
net: dsa: b53: fix an off by one in checking "vlan->vid"
tcp: do not mess with cloned skbs in tcp_add_backlog()
selftests: net: fib_tests: remove duplicate log test
net: nfc: nci: fix the wrong NCI_CORE_INIT parameters
sh_eth: Fix power down vs. is_opened flag ordering
net: Disable NETIF_F_HW_TLS_RX when RXCSUM is disabled
netfilter: rpfilter: mask ecn bits before fib lookup
udp: mask TOS bits in udp_v4_early_demux()
xsk: Clear pool even for inactive queues
bpf: Fix helper bpf_map_peek_elem_proto pointing to wrong callback
sh_eth: Make PHY access aware of Runtime PM to fix reboot crash
...
Fix incorrect signed_{sub,add32}_overflows() input types (and a related buggy
comment). It looks like this might have slipped in via copy/paste issue, also
given prior to 3f50f132d8 ("bpf: Verifier, do explicit ALU32 bounds tracking")
the signature of signed_sub_overflows() had s64 a and s64 b as its input args
whereas now they are truncated to s32. Thus restore proper types. Also, the case
of signed_add32_overflows() is not consistent to signed_sub32_overflows(). Both
have s32 as inputs, therefore align the former.
Fixes: 3f50f132d8 ("bpf: Verifier, do explicit ALU32 bounds tracking")
Reported-by: De4dCr0w <sa516203@mail.ustc.edu.cn>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Pull task_work fix from Jens Axboe:
"The TIF_NOTIFY_SIGNAL change inadvertently removed the unconditional
task_work run we had in get_signal().
This caused a regression for some setups, since we're relying on eg
____fput() being run to close and release, for example, a pipe and
wake the other end.
For 5.11, I prefer the simple solution of just reinstating the
unconditional run, even if it conceptually doesn't make much sense -
if you need that kind of guarantee, you should be using TWA_SIGNAL
instead of TWA_NOTIFY. But it's the trivial fix for 5.11, and would
ensure that other potential gotchas/assumptions for task_work don't
regress for 5.11.
We're looking into further simplifying the task_work notifications for
5.12 which would resolve that too"
* tag 'task_work-2021-01-19' of git://git.kernel.dk/linux-block:
task_work: unconditionally run task_work from get_signal()
Fix NULL pointer dereference when adding new psi monitor to the root
cgroup. PSI files for root cgroup was introduced in df5ba5be74 by using
system wide psi struct when reading, but file write/monitor was not
properly fixed. Since the PSI config for the root cgroup isn't
initialized, the current implementation tries to lock a NULL ptr,
resulting in a crash.
Can be triggered by running this as root:
$ tee /sys/fs/cgroup/cpu.pressure <<< "some 10000 1000000"
Signed-off-by: Odin Ugedal <odin@uged.al>
Reviewed-by: Suren Baghdasaryan <surenb@google.com>
Acked-by: Dan Schatzberg <dschatzberg@fb.com>
Fixes: df5ba5be74 ("kernel/sched/psi.c: expose pressure metrics on root cgroup")
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: stable@vger.kernel.org # 5.2+
Signed-off-by: Tejun Heo <tj@kernel.org>
Before the commit 896fbe20b4 ("printk: use the lockless
ringbuffer"), msg_print_text() would only write up to size-1 bytes
into the provided buffer. Some callers expect this behavior and
append a terminator to returned string. In particular:
arch/powerpc/xmon/xmon.c:dump_log_buf()
arch/um/kernel/kmsg_dump.c:kmsg_dumper_stdout()
msg_print_text() has been replaced by record_print_text(), which
currently fills the full size of the buffer. This causes a
buffer overflow for the above callers.
Change record_print_text() so that it will only use size-1 bytes
for text data. Also, for paranoia sakes, add a terminator after
the text data.
And finally, document this behavior so that it is clear that only
size-1 bytes are used and a terminator is added.
Fixes: 896fbe20b4 ("printk: use the lockless ringbuffer")
Cc: stable@vger.kernel.org # 5.10+
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Acked-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20210114170412.4819-1-john.ogness@linutronix.de
5fdc7db644 ("module: setup load info before module_sig_check()")
moved the ELF setup, so that it was done before the signature
check. This made the module name available to signature error
messages.
However, the checks for ELF correctness in setup_load_info
are not sufficient to prevent bad memory references due to
corrupted offset fields, indices, etc.
So, there's a regression in behavior here: a corrupt and unsigned
(or badly signed) module, which might previously have been rejected
immediately, can now cause an oops/crash.
Harden ELF handling for module loading by doing the following:
- Move the signature check back up so that it comes before ELF
initialization. It's best to do the signature check to see
if we can trust the module, before using the ELF structures
inside it. This also makes checks against info->len
more accurate again, as this field will be reduced by the
length of the signature in mod_check_sig().
The module name is now once again not available for error
messages during the signature check, but that seems like
a fair tradeoff.
- Check if sections have offset / size fields that at least don't
exceed the length of the module.
- Check if sections have section name offsets that don't fall
outside the section name table.
- Add a few other sanity checks against invalid section indices,
etc.
This is not an exhaustive consistency check, but the idea is to
at least get through the signature and blacklist checks without
crashing because of corrupted ELF info, and to error out gracefully
for most issues that would have caused problems later on.
Fixes: 5fdc7db644 ("module: setup load info before module_sig_check()")
Signed-off-by: Frank van der Linden <fllinden@amazon.com>
Signed-off-by: Jessica Yu <jeyu@kernel.org>
clang-12 -fno-pic (since
a084c0388e)
can emit `call __stack_chk_fail@PLT` instead of `call __stack_chk_fail`
on x86. The two forms should have identical behaviors on x86-64 but the
former causes GNU as<2.37 to produce an unreferenced undefined symbol
_GLOBAL_OFFSET_TABLE_.
(On x86-32, there is an R_386_PC32 vs R_386_PLT32 difference but the
linker behavior is identical as far as Linux kernel is concerned.)
Simply ignore _GLOBAL_OFFSET_TABLE_ for now, like what
scripts/mod/modpost.c:ignore_undef_symbol does. This also fixes the
problem for gcc/clang -fpie and -fpic, which may emit `call foo@PLT` for
external function calls on x86.
Note: ld -z defs and dynamic loaders do not error for unreferenced
undefined symbols so the module loader is reading too much. If we ever
need to ignore more symbols, the code should be refactored to ignore
unreferenced symbols.
Cc: <stable@vger.kernel.org>
Link: https://github.com/ClangBuiltLinux/linux/issues/1250
Link: https://sourceware.org/bugzilla/show_bug.cgi?id=27178
Reported-by: Marco Elver <elver@google.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Marco Elver <elver@google.com>
Signed-off-by: Fangrui Song <maskray@google.com>
Signed-off-by: Jessica Yu <jeyu@kernel.org>
Daniel Borkmann says:
====================
pull-request: bpf-next 2021-01-16
1) Extend atomic operations to the BPF instruction set along with x86-64 JIT support,
that is, atomic{,64}_{xchg,cmpxchg,fetch_{add,and,or,xor}}, from Brendan Jackman.
2) Add support for using kernel module global variables (__ksym externs in BPF
programs) retrieved via module's BTF, from Andrii Nakryiko.
3) Generalize BPF stackmap's buildid retrieval and add support to have buildid
stored in mmap2 event for perf, from Jiri Olsa.
4) Various fixes for cross-building BPF sefltests out-of-tree which then will
unblock wider automated testing on ARM hardware, from Jean-Philippe Brucker.
5) Allow to retrieve SOL_SOCKET opts from sock_addr progs, from Daniel Borkmann.
6) Clean up driver's XDP buffer init and split into two helpers to init per-
descriptor and non-changing fields during processing, from Lorenzo Bianconi.
7) Minor misc improvements to libbpf & bpftool, from Ian Rogers.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (41 commits)
perf: Add build id data in mmap2 event
bpf: Add size arg to build_id_parse function
bpf: Move stack_map_get_build_id into lib
bpf: Document new atomic instructions
bpf: Add tests for new BPF atomic operations
bpf: Add bitwise atomic instructions
bpf: Pull out a macro for interpreting atomic ALU operations
bpf: Add instructions for atomic_[cmp]xchg
bpf: Add BPF_FETCH field / create atomic_fetch_add instruction
bpf: Move BPF_STX reserved field check into BPF_STX verifier code
bpf: Rename BPF_XADD and prepare to encode other atomics in .imm
bpf: x86: Factor out a lookup table for some ALU opcodes
bpf: x86: Factor out emission of REX byte
bpf: x86: Factor out emission of ModR/M for *(reg + off)
tools/bpftool: Add -Wall when building BPF programs
bpf, libbpf: Avoid unused function warning on bpf_tail_call_static
selftests/bpf: Install btf_dump test cases
selftests/bpf: Fix installation of urandom_read
selftests/bpf: Move generated test files to $(TEST_GEN_FILES)
selftests/bpf: Fix out-of-tree build
...
====================
Link: https://lore.kernel.org/r/20210116012922.17823-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Daniel Borkmann says:
====================
pull-request: bpf 2021-01-16
1) Fix a double bpf_prog_put() for BPF_PROG_{TYPE_EXT,TYPE_TRACING} types in
link creation's error path causing a refcount underflow, from Jiri Olsa.
2) Fix BTF validation errors for the case where kernel modules don't declare
any new types and end up with an empty BTF, from Andrii Nakryiko.
3) Fix BPF local storage helpers to first check their {task,inode} owners for
being NULL before access, from KP Singh.
4) Fix a memory leak in BPF setsockopt handling for the case where optlen is
zero and thus temporary optval buffer should be freed, from Stanislav Fomichev.
5) Fix a syzbot memory allocation splat in BPF_PROG_TEST_RUN infra for
raw_tracepoint caused by too big ctx_size_in, from Song Liu.
6) Fix LLVM code generation issues with verifier where PTR_TO_MEM{,_OR_NULL}
registers were spilled to stack but not recognized, from Gilad Reti.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
MAINTAINERS: Update my email address
selftests/bpf: Add verifier test for PTR_TO_MEM spill
bpf: Support PTR_TO_MEM{,_OR_NULL} register spilling
bpf: Reject too big ctx_size_in for raw_tp test run
libbpf: Allow loading empty BTFs
bpf: Allow empty module BTFs
bpf: Don't leak memory in bpf getsockopt when optlen == 0
bpf: Update local storage test to check handling of null ptrs
bpf: Fix typo in bpf_inode_storage.c
bpf: Local storage helpers should check nullness of owner ptr passed
bpf: Prevent double bpf_prog_put call from bpf_tracing_prog_attach
====================
Link: https://lore.kernel.org/r/20210116002025.15706-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The functions cgroup_threads_write and cgroup_procs_write are almost
identical. In order to reduce duplication, factor out the common code in
similar fashion we already do for other threadgroup/task functions. No
functional changes are intended.
Suggested-by: Hao Lee <haolee.swjtu@gmail.com>
Signed-off-by: Michal Koutný <mkoutny@suse.com>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
When mounting a cgroup hierarchy with disabled controller in cgroup v1,
all available controllers will be attached.
For example, boot with cgroup_no_v1=cpu or cgroup_disable=cpu, and then
mount with "mount -t cgroup -ocpu cpu /sys/fs/cgroup/cpu", then all
enabled controllers will be attached except cpu.
Fix this by adding disabled controller check in cgroup1_parse_param().
If the specified controller is disabled, just return error with information
"Disabled controller xx" rather than attaching all the other enabled
controllers.
Fixes: f5dfb5315d ("cgroup: take options parsing into ->parse_monolithic()")
Signed-off-by: Chen Zhou <chenzhou10@huawei.com>
Reviewed-by: Zefan Li <lizefan.x@bytedance.com>
Reviewed-by: Michal Koutný <mkoutny@suse.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
This was replaced with a kauditd_wait kthread long ago,
back in:
b7d1125817 (AUDIT: Send netlink messages from a separate kernel thread)
Update the stale comment.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul Moore <paul@paul-moore.com>
kmsg_dump_get_buffer() uses @syslog to determine if the syslog
prefix should be written to the buffer. However, when calculating
the maximum number of records that can fit into the buffer, it
always counts the bytes from the syslog prefix.
Use @syslog when calculating the maximum number of records that can
fit into the buffer.
Fixes: e2ae715d66 ("kmsg - kmsg_dump() use iterator to receive log buffer content")
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Acked-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20210113164413.1599-1-john.ogness@linutronix.de
Counting text lines in a record simply involves counting the number
of newline characters (+1). However, it is searching the full data
block for newline characters, even though the text data can be (and
often is) a subset of that area. Since the extra area in the data
block was never initialized, the result is that extra newlines may
be seen and counted.
Restrict newline searching to the text data length.
Fixes: b6cf8b3f33 ("printk: add lockless ringbuffer")
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Acked-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20210113144234.6545-1-john.ogness@linutronix.de
Adding support to carry build id data in mmap2 event.
The build id data replaces maj/min/ino/ino_generation
fields, which are also used to identify map's binary,
so it's ok to replace them with build id data:
union {
struct {
u32 maj;
u32 min;
u64 ino;
u64 ino_generation;
};
struct {
u8 build_id_size;
u8 __reserved_1;
u16 __reserved_2;
u8 build_id[20];
};
};
Replaced maj/min/ino/ino_generation fields give us size
of 24 bytes. We use 20 bytes for build id data, 1 byte
for size and rest is unused.
There's new misc bit for mmap2 to signal there's build
id data in it:
#define PERF_RECORD_MISC_MMAP_BUILD_ID (1 << 14)
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/bpf/20210114134044.1418404-4-jolsa@kernel.org
It's possible to have other build id types (other than default SHA1).
Currently there's also ld support for MD5 build id.
Adding size argument to build_id_parse function, that returns (if defined)
size of the parsed build id, so we can recognize the build id type.
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210114134044.1418404-3-jolsa@kernel.org
Moving stack_map_get_build_id into lib with
declaration in linux/buildid.h header:
int build_id_parse(struct vm_area_struct *vma, unsigned char *build_id);
This function returns build id for given struct vm_area_struct.
There is no functional change to stack_map_get_build_id function.
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Link: https://lore.kernel.org/bpf/20210114134044.1418404-2-jolsa@kernel.org
This adds instructions for
atomic[64]_[fetch_]and
atomic[64]_[fetch_]or
atomic[64]_[fetch_]xor
All these operations are isomorphic enough to implement with the same
verifier, interpreter, and x86 JIT code, hence being a single commit.
The main interesting thing here is that x86 doesn't directly support
the fetch_ version these operations, so we need to generate a CMPXCHG
loop in the JIT. This requires the use of two temporary registers,
IIUC it's safe to use BPF_REG_AX and x86's AUX_REG for this purpose.
Signed-off-by: Brendan Jackman <jackmanb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20210114181751.768687-10-jackmanb@google.com
Since the atomic operations that are added in subsequent commits are
all isomorphic with BPF_ADD, pull out a macro to avoid the
interpreter becoming dominated by lines of atomic-related code.
Note that this sacrificies interpreter performance (combining
STX_ATOMIC_W and STX_ATOMIC_DW into single switch case means that we
need an extra conditional branch to differentiate them) in favour of
compact and (relatively!) simple C code.
Signed-off-by: Brendan Jackman <jackmanb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20210114181751.768687-9-jackmanb@google.com
This adds two atomic opcodes, both of which include the BPF_FETCH
flag. XCHG without the BPF_FETCH flag would naturally encode
atomic_set. This is not supported because it would be of limited
value to userspace (it doesn't imply any barriers). CMPXCHG without
BPF_FETCH woulud be an atomic compare-and-write. We don't have such
an operation in the kernel so it isn't provided to BPF either.
There are two significant design decisions made for the CMPXCHG
instruction:
- To solve the issue that this operation fundamentally has 3
operands, but we only have two register fields. Therefore the
operand we compare against (the kernel's API calls it 'old') is
hard-coded to be R0. x86 has similar design (and A64 doesn't
have this problem).
A potential alternative might be to encode the other operand's
register number in the immediate field.
- The kernel's atomic_cmpxchg returns the old value, while the C11
userspace APIs return a boolean indicating the comparison
result. Which should BPF do? A64 returns the old value. x86 returns
the old value in the hard-coded register (and also sets a
flag). That means return-old-value is easier to JIT, so that's
what we use.
Signed-off-by: Brendan Jackman <jackmanb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20210114181751.768687-8-jackmanb@google.com
A subsequent patch will add additional atomic operations. These new
operations will use the same opcode field as the existing XADD, with
the immediate discriminating different operations.
In preparation, rename the instruction mode BPF_ATOMIC and start
calling the zero immediate BPF_ADD.
This is possible (doesn't break existing valid BPF progs) because the
immediate field is currently reserved MBZ and BPF_ADD is zero.
All uses are removed from the tree but the BPF_XADD definition is
kept around to avoid breaking builds for people including kernel
headers.
Signed-off-by: Brendan Jackman <jackmanb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Björn Töpel <bjorn.topel@gmail.com>
Link: https://lore.kernel.org/bpf/20210114181751.768687-5-jackmanb@google.com
The purpose of local_lock_t is to abstract: preempt_disable() /
local_bh_disable() / local_irq_disable(). These are the traditional
means of gaining access to per-cpu data, but are fundamentally
non-preemptible.
local_lock_t provides a per-cpu lock, that on !PREEMPT_RT reduces to
no-ops, just like regular spinlocks do on UP.
This gives rise to:
CPU0 CPU1
local_lock(B) spin_lock_irq(A)
<IRQ>
spin_lock(A) local_lock(B)
Where lockdep then figures things will lock up; which would be true if
B were any other kind of lock. However this is a false positive, no
such deadlock actually exists.
For !RT the above local_lock(B) is preempt_disable(), and there's
obviously no deadlock; alternatively, CPU0's B != CPU1's B.
For RT the argument is that since local_lock() nests inside
spin_lock(), it cannot be used in hardirq context, and therefore CPU0
cannot in fact happen. Even though B is a real lock, it is a
preemptible lock and any threaded-irq would simply schedule out and
let the preempted task (which holds B) continue such that the task on
CPU1 can make progress, after which the threaded-irq resumes and can
finish.
This means that we can never form an IRQ inversion on a local_lock
dependency, so terminate the graph walk when looking for IRQ
inversions when we encounter one.
One consequence is that (for LOCKDEP_SMALL) when we look for redundant
dependencies, A -> B is not redundant in the presence of A -> L -> B.
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
[peterz: Changelog]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
In preparation for adding an TRACE_IRQFLAGS dependent skip function to
check_redundant(), move it below the TRACE_IRQFLAGS #ifdef.
While there, provide a stub function to reduce #ifdef usage.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Some __bfs() walks will have additional iteration constraints (beyond
the path being strong). Provide an additional function to allow
terminating graph walks.
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
The local_lock_t's are special, because they cannot form IRQ
inversions, make sure we can tell them apart from the rest of the
locks.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Active balance is triggered for a number of voluntary cases like misfit
or pinned tasks cases but also after that a number of load balance
attempts failed to migrate a task. There is no need to use active load
balance when the group is overloaded because an overloaded state means
that there is at least one waiting task. Nevertheless, the waiting task
is not selected and detached until the threshold becomes higher than its
load. This threshold increases with the number of failed lb (see the
condition if ((load >> env->sd->nr_balance_failed) > env->imbalance) in
detach_tasks()) and the waiting task will end up to be selected after a
number of attempts.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lkml.kernel.org/r/20210107103325.30851-4-vincent.guittot@linaro.org
Setting LBF_ALL_PINNED during active load balance is only valid when there
is only 1 running task on the rq otherwise this ends up increasing the
balance interval whereas other tasks could migrate after the next interval
once they become cache-cold as an example.
LBF_ALL_PINNED flag is now always set it by default. It is then cleared
when we find one task that can be pulled when calling detach_tasks() or
during active migration.
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lkml.kernel.org/r/20210107103325.30851-3-vincent.guittot@linaro.org
CPU (root cfs_rq) estimated utilization (util_est) is currently used in
dequeue_task_fair() to drive frequency selection before it is updated.
with:
CPU_util : rq->cfs.avg.util_avg
CPU_util_est : rq->cfs.avg.util_est
CPU_utilization : max(CPU_util, CPU_util_est)
task_util : p->se.avg.util_avg
task_util_est : p->se.avg.util_est
dequeue_task_fair():
/* (1) CPU_util and task_util update + inform schedutil about
CPU_utilization changes */
for_each_sched_entity() /* 2 loops */
(dequeue_entity() ->) update_load_avg() -> cfs_rq_util_change()
-> cpufreq_update_util() ->...-> sugov_update_[shared\|single]
-> sugov_get_util() -> cpu_util_cfs()
/* (2) CPU_util_est and task_util_est update */
util_est_dequeue()
cpu_util_cfs() uses CPU_utilization which could lead to a false (too
high) utilization value for schedutil in task ramp-down or ramp-up
scenarios during task dequeue.
To mitigate the issue split the util_est update (2) into:
(A) CPU_util_est update in util_est_dequeue()
(B) task_util_est update in util_est_update()
Place (A) before (1) and keep (B) where (2) is. The latter is necessary
since (B) relies on task_util update in (1).
Fixes: 7f65ea42eb ("sched/fair: Add util_est on top of PELT")
Signed-off-by: Xuewen Yan <xuewen.yan@unisoc.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/1608283672-18240-1-git-send-email-xuewen.yan94@gmail.com
SCHED_SOFTIRQ is raised to trigger periodic load balancing. When CPU is not
active, CPU should not participate in load balancing.
The scheduler uses nohz.idle_cpus_mask to keep track of the CPUs which can
do idle load balancing. When bringing a CPU up the CPU is added to the mask
when it reaches the active state, but on teardown the CPU stays in the mask
until it goes offline and invokes sched_cpu_dying().
When SCHED_SOFTIRQ is raised on a !active CPU, there might be a pending
softirq when stopping the tick which triggers a warning in NOHZ code. The
SCHED_SOFTIRQ can also be raised by the scheduler tick which has the same
issue.
Therefore remove the CPU from nohz.idle_cpus_mask when it is marked
inactive and also prevent the scheduler_tick() from raising SCHED_SOFTIRQ
after this point.
Signed-off-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Link: https://lkml.kernel.org/r/20201215104400.9435-1-anna-maria@linutronix.de
Add support for pointer to mem register spilling, to allow the verifier
to track pointers to valid memory addresses. Such pointers are returned
for example by a successful call of the bpf_ringbuf_reserve helper.
The patch was partially contributed by CyberArk Software, Inc.
Fixes: 457f44363a ("bpf: Implement BPF ring buffer and verifier support for it")
Suggested-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Gilad Reti <gilad.reti@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: KP Singh <kpsingh@kernel.org>
Link: https://lore.kernel.org/bpf/20210113053810.13518-1-gilad.reti@gmail.com
Add support for directly accessing kernel module variables from BPF programs
using special ldimm64 instructions. This functionality builds upon vmlinux
ksym support, but extends ldimm64 with src_reg=BPF_PSEUDO_BTF_ID to allow
specifying kernel module BTF's FD in insn[1].imm field.
During BPF program load time, verifier will resolve FD to BTF object and will
take reference on BTF object itself and, for module BTFs, corresponding module
as well, to make sure it won't be unloaded from under running BPF program. The
mechanism used is similar to how bpf_prog keeps track of used bpf_maps.
One interesting change is also in how per-CPU variable is determined. The
logic is to find .data..percpu data section in provided BTF, but both vmlinux
and module each have their own .data..percpu entries in BTF. So for module's
case, the search for DATASEC record needs to look at only module's added BTF
types. This is implemented with custom search function.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Acked-by: Hao Luo <haoluo@google.com>
Link: https://lore.kernel.org/bpf/20210112075520.4103414-6-andrii@kernel.org
Pull irqchip fixes from Marc Zyngier:
- Fix the MIPS CPU interrupt controller hierarchy
- Simplify the PRUSS Kconfig entry
- Eliminate trivial build warnings on the MIPS Loongson liointc
- Fix error path in devm_platform_get_irqs_affinity()
- Turn the BCM2836 IPI irq_eoi callback into irq_ack
- Fix initialisation of on-stack msi_alloc_info
- Cleanup spurious comma in irq-sl28cpld
Link: https://lore.kernel.org/r/20210110110001.2328708-1-maz@kernel.org
Due to an integer overflow, RTC synchronization now happens every 2s
instead of the intended 11 minutes. Fix this by forcing 64-bit
arithmetic for the sync period calculation.
Annotate the other place which multiplies seconds for consistency as well.
Fixes: c9e6189fb0 ("ntp: Make the RTC synchronization more reliable")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20210111103956.290378-1-geert+renesas@glider.be
