Commit b33fff07e3 ("x86, build: allow LTO to be selected") added a
couple of '-plugin-opt=' flags to KBUILD_LDFLAGS because the code model
and stack alignment are not stored in LLVM bitcode.
However, these flags were added to KBUILD_LDFLAGS prior to the
emulation flag assignment, which uses ':=', so they were overwritten
and never added to $(LD) invocations.
The absence of these flags caused misalignment issues in the
AMDGPU driver when compiling with CONFIG_LTO_CLANG, resulting in
general protection faults.
Shuffle the assignment below the initial one so that the flags are
properly passed along and all of the linker flags stay together.
At the same time, avoid any future issues with clobbering flags by
changing the emulation flag assignment to '+=' since KBUILD_LDFLAGS is
already defined with ':=' in the main Makefile before being exported for
modification here as a result of commit:
ce99d0bf31 ("kbuild: clear LDFLAGS in the top Makefile")
Fixes: b33fff07e3 ("x86, build: allow LTO to be selected")
Reported-by: Anthony Ruhier <aruhier@mailbox.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Anthony Ruhier <aruhier@mailbox.org>
Cc: stable@vger.kernel.org
Link: https://github.com/ClangBuiltLinux/linux/issues/1374
Link: https://lore.kernel.org/r/20210518190106.60935-1-nathan@kernel.org
The __nvmf_check_ready() routine used to bounce all filesystem io if the
controller state isn't LIVE. However, a later patch changed the logic so
that it rejection ends up being based on the Q live check. The FC
transport has a slightly different sequence from rdma and tcp for
shutting down queues/marking them non-live. FC marks its queue non-live
after aborting all ios and waiting for their termination, leaving a
rather large window for filesystem io to continue to hit the transport.
Unfortunately this resulted in filesystem I/O or applications seeing I/O
errors.
Change the FC transport to mark the queues non-live at the first sign of
teardown for the association (when I/O is initially terminated).
Fixes: 73a5379937 ("nvme-fabrics: allow to queue requests for live queues")
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
A possible race condition exists where the request to send data is
enqueued from nvme_tcp_handle_r2t()'s will not be observed by
nvme_tcp_send_all() if it happens to be running. The driver relies on
io_work to send the enqueued request when it is runs again, but the
concurrently running nvme_tcp_send_all() may not have released the
send_mutex at that time. If no future commands are enqueued to re-kick
the io_work, the request will timeout in the SEND_H2C state, resulting
in a timeout error like:
nvme nvme0: queue 1: timeout request 0x3 type 6
Ensure the io_work continues to run as long as the req_list is not empty.
Fixes: db5ad6b7f8 ("nvme-tcp: try to send request in queue_rq context")
Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Commit db5ad6b7f8 ("nvme-tcp: try to send request in queue_rq
context") added a second context that may perform a network send.
This means that now RX and TX are not serialized in nvme_tcp_io_work
and can run concurrently.
While there is correct mutual exclusion in the TX path (where
the send_mutex protect the queue socket send activity) RX activity,
and more specifically request completion may run concurrently.
This means we must guarantee that any mutation of the request state
related to its lifetime, bytes sent must not be accessed when a completion
may have possibly arrived back (and processed).
The race may trigger when a request completion arrives, processed
_and_ reused as a fresh new request, exactly in the (relatively short)
window between the last data payload sent and before the request iov_iter
is advanced.
Consider the following race:
1. 16K write request is queued
2. The nvme command and the data is sent to the controller (in-capsule
or solicited by r2t)
3. After the last payload is sent but before the req.iter is advanced,
the controller sends back a completion.
4. The completion is processed, the request is completed, and reused
to transfer a new request (write or read)
5. The new request is queued, and the driver reset the request parameters
(nvme_tcp_setup_cmd_pdu).
6. Now context in (2) resumes execution and advances the req.iter
==> use-after-completion as this is already a new request.
Fix this by making sure the request is not advanced after the last
data payload send, knowing that a completion may have arrived already.
An alternative solution would have been to delay the request completion
or state change waiting for reference counting on the TX path, but besides
adding atomic operations to the hot-path, it may present challenges in
multi-stage R2T scenarios where a r2t handler needs to be deferred to
an async execution.
Reported-by: Narayan Ayalasomayajula <narayan.ayalasomayajula@wdc.com>
Tested-by: Anil Mishra <anil.mishra@wdc.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Cc: stable@vger.kernel.org # v5.8+
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When creating loop ctrl in nvme_loop_create_ctrl(), if nvme_init_ctrl()
fails, the loop ctrl should be freed before jumping to the "out" label.
Fixes: 3a85a5de29 ("nvme-loop: add a NVMe loopback host driver")
Signed-off-by: Wu Bo <wubo40@huawei.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When creating ctrl in nvmet_alloc_ctrl(), if the cntlid_min is larger
than cntlid_max of the subsystem, and jumps to the
"out_free_changed_ns_list" label, but the ctrl->sqs lack of be freed.
Fix this by jumping to the "out_free_sqs" label.
Fixes: 94a39d61f8 ("nvmet: make ctrl-id configurable")
Signed-off-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
It's not correct to call napi_schedule() in pure process
context. Because we use __raise_softirq_irqoff() we require
callers to be in a context which will eventually lead to
softirq handling (hardirq, bh disabled, etc.).
With code as is users will see:
NOHZ tick-stop error: Non-RCU local softirq work is pending, handler #08!!!
Fixes: a8dd7ac12f ("net/mlx5e: Generalize RQ activation")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Termination tables are restricted to have the default miss action and
cannot be set to forward to another table in case of a miss.
If the fs prio of the termination table is not the last one in the
list, fs_core will attempt to attach it to another table.
Set the unmanaged ft flag when creating the termination table ft
and select the tc offload prio for it to prevent fs_core from selecting
the forwarding to next ft miss action and use the default one.
In addition, set the flow that forwards to the termination table to
ignore ft level restrictions since the ft level is not set by fs_core
for unamanged fts.
Fixes: 249ccc3c95 ("net/mlx5e: Add support for offloading traffic from uplink to uplink")
Signed-off-by: Ariel Levkovich <lariel@nvidia.com>
During driver probe of device that has dynamic MSI-X feature enabled,
the following error is printed in some FW flavour (not released yet).
mlx5_core 0000:06:00.0: firmware version: 4.7.4387
mlx5_core 0000:06:00.0: 126.016 Gb/s available PCIe bandwidth (8.0 GT/s PCIe x16 link)
mlx5_core 0000:06:00.0: mlx5_cmd_check:777:(pid 70599): SET_HCA_CAP(0x109) op_mod(0x0) failed, status bad parameter(0x3), syndrome (0x0)
mlx5_core 0000:06:00.0: set_hca_cap:622:(pid 70599): handle_hca_cap failed
mlx5_core 0000:06:00.0: mlx5_function_setup:1045:(pid 70599): set_hca_cap failed
mlx5_core 0000:06:00.0: probe_one:1465:(pid 70599): mlx5_init_one failed with error code -22
mlx5_core: probe of 0000:06:00.0 failed with error -22
In order to make the setting capability of MSI-X future proof, let's
query the current capabilities first.
Fixes: 604774add5 ("net/mlx5: Dynamically assign MSI-X vectors count")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
net/mlx5: Expose MPFS configuration API
MPFS is the multi physical function switch that bridges traffic between
the physical port and any physical functions associated with it. The
driver is required to add or remove MAC entries to properly forward
incoming traffic to the correct physical function.
We export the API to control MPFS so that other drivers, such as
mlx5_vdpa are able to add MAC addresses of their network interfaces.
The MAC address of the vdpa interface must be configured into the MPFS L2
address. Failing to do so could cause, in some NIC configurations, failure
to forward packets to the vdpa network device instance.
Fix this by adding calls to update the MPFS table.
CC: <mst@redhat.com>
CC: <jasowang@redhat.com>
CC: <virtualization@lists.linux-foundation.org>
Fixes: 1a86b377aa ("vdpa/mlx5: Add VDPA driver for supported mlx5 devices")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Avoid division by zero in the error flow. In the driver TC number can be
either 1 or 8. When TC count is set to 1, driver zero netdev->num_tc.
Hence, need to convert it back from 0 to 1 in the error flow.
Fixes: fa3748775b ("net/mlx5e: Handle errors from netif_set_real_num_{tx,rx}_queues")
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Rules with MLX5_ESW_DEST_CHAIN_WITH_SRC_PORT_CHANGE dest flag are
translated to destination FT in eswitch. Currently it is not possible to
mirror such rules because firmware doesn't support mixing FT and Vport
destinations in single rule when one of them adds encapsulation. Since the
only use case for MLX5_ESW_DEST_CHAIN_WITH_SRC_PORT_CHANGE destination is
support for tunnel endpoints on VF and trying to offload such rule with
mirror action causes either crash in fs_core or firmware error with
syndrome 0xff6a1d, reject all such rules in mlx5 TC layer.
Fixes: 10742efc20 ("net/mlx5e: VF tunnel TX traffic offloading")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
When handling FIB_EVENT_ENTRY_REPLACE event for a new multipath route,
lag activation can be missed if a stale (struct lag_mp)->mfi pointer
exists, which was associated with an older multipath route that had been
removed.
Normally, when a route is removed, it triggers mlx5_lag_fib_event(),
which handles FIB_EVENT_ENTRY_DEL and clears mfi pointer. But, if
mlx5_lag_check_prereq() condition isn't met, for example when eswitch is
in legacy mode, the fib event is skipped and mfi pointer becomes stale.
Fix by resetting mfi pointer to NULL every time mlx5_lag_mp_init() is
called.
Fixes: 544fe7c2e6 ("net/mlx5e: Activate HW multipath and handle port affinity based on FIB events")
Signed-off-by: Dima Chumak <dchumak@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
mlx5e_attach_netdev can be called prior to registering the netdevice:
Example stack:
ipoib_new_child_link ->
ipoib_intf_init->
rdma_init_netdev->
mlx5_rdma_setup_rn->
mlx5e_attach_netdev->
mlx5e_num_channels_changed ->
mlx5e_set_default_xps_cpumasks ->
netif_set_xps_queue ->
__netif_set_xps_queue -> kmalloc
If any later stage fails at any point after mlx5e_num_channels_changed()
returns, XPS allocated maps will never be freed as they
are only freed during netdev unregistration, which will never happen for
yet to be registered netdevs.
Fixes: 3909a12e79 ("net/mlx5e: Fix configuration of XPS cpumasks and netdev queues in corner cases")
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Aya Levin <ayal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
It could be the lag dev is null so stop processing the event.
In bond_enslave() the active/backup slave being set before setting the
upper dev so first event is without an upper dev.
After setting the upper dev with bond_master_upper_dev_link() there is
a second event and in that event we have an upper dev.
Fixes: 7e51891a23 ("net/mlx5e: Use netdev events to set/del egress acl forward-to-vport rule")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The result of __dev_get_by_index() is not checked for NULL, which then
passed to mlx5e_attach_encap() and gets dereferenced.
Also, in case of a successful lookup, the net_device reference count is
not incremented, which may result in net_device pointer becoming invalid
at any time during mlx5e_attach_encap() execution.
Fix by using dev_get_by_index(), which does proper reference counting on
the net_device pointer. Also, handle nullptr return value when mirred
device is not found.
It's safe to call dev_put() on the mirred net_device pointer, right
after mlx5e_attach_encap() call, because it's not being saved/copied
down the call chain.
Fixes: 3c37745ec6 ("net/mlx5e: Properly deal with encap flows add/del under neigh update")
Addresses-Coverity: ("Dereference null return value")
Signed-off-by: Dima Chumak <dchumak@nvidia.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
When a SF is inactivated and when it is in a TEARDOWN_REQUEST
state, driver still returns its state as active. This is incorrect.
Fix it by treating TEARDOWN_REQEUST as inactive state. When a SF
is still attached to the driver, on user request to reactivate EINVAL
error is returned. Inform user about it with better code EBUSY and
informative error message.
Fixes: 6a32732174 ("net/mlx5: SF, Port function state change support")
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Vu Pham <vuhuong@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Fix print to print correct error code and not using IS_ERR() which
will just result in always printing 1.
Also return real err instead of always -EOPNOTSUPP.
Fixes: 10caabdaad ("net/mlx5e: Use termination table for VLAN push actions")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
For remote mirroring, after the tunnel packets are received, they are
decapsulated and sent to representor, then re-encapsulated and sent
out over another tunnel. So reformat action is set only when the
destination is required to do encapsulation.
Fixes: 249ccc3c95 ("net/mlx5e: Add support for offloading traffic from uplink to uplink")
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Ariel Levkovich <lariel@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
The result of dev_get_by_index_rcu() is not checked for NULL and then
gets dereferenced immediately.
Also, the RCU lock must be held by the caller of dev_get_by_index_rcu(),
which isn't satisfied by the call stack.
Fix by handling nullptr return value when iflink device is not found.
Add RCU locking around dev_get_by_index_rcu() to avoid possible adverse
effects while iterating over the net_device's hlist.
It is safe not to increment reference count of the net_device pointer in
case of a successful lookup, because it's already handled by VLAN code
during VLAN device registration (see register_vlan_dev and
netdev_upper_dev_link).
Fixes: 278748a95a ("net/mlx5e: Offload TC e-switch rules with egress VLAN device")
Addresses-Coverity: ("Dereference null return value")
Signed-off-by: Dima Chumak <dchumak@nvidia.com>
Reviewed-by: Vlad Buslov <vladbu@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
mlx5_core_dev holds pointer to static profile, hence when the
log_max_qp of the profile is override by some device, then it
effect all other mlx5 devices that share the same profile.
Fix it by having a profile instance for every mlx5 device.
Fixes: 883371c453 ("net/mlx5: Check FW limitations on log_max_qp before setting it")
Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Add -L flag to bpftool to use libbpf gen_trace facility and syscall/loader program
for skeleton generation and program loading.
"bpftool gen skeleton -L" command will generate a "light skeleton" or "loader skeleton"
that is similar to existing skeleton, but has one major difference:
$ bpftool gen skeleton lsm.o > lsm.skel.h
$ bpftool gen skeleton -L lsm.o > lsm.lskel.h
$ diff lsm.skel.h lsm.lskel.h
@@ -5,34 +4,34 @@
#define __LSM_SKEL_H__
#include <stdlib.h>
-#include <bpf/libbpf.h>
+#include <bpf/bpf.h>
The light skeleton does not use majority of libbpf infrastructure.
It doesn't need libelf. It doesn't parse .o file.
It only needs few sys_bpf wrappers. All of them are in bpf/bpf.h file.
In future libbpf/bpf.c can be inlined into bpf.h, so not even libbpf.a would be
needed to work with light skeleton.
"bpftool prog load -L file.o" command is introduced for debugging of syscall/loader
program generation. Just like the same command without -L it will try to load
the programs from file.o into the kernel. It won't even try to pin them.
"bpftool prog load -L -d file.o" command will provide additional debug messages
on how syscall/loader program was generated.
Also the execution of syscall/loader program will use bpf_trace_printk() for
each step of loading BTF, creating maps, and loading programs.
The user can do "cat /.../trace_pipe" for further debug.
An example of fexit_sleep.lskel.h generated from progs/fexit_sleep.c:
struct fexit_sleep {
struct bpf_loader_ctx ctx;
struct {
struct bpf_map_desc bss;
} maps;
struct {
struct bpf_prog_desc nanosleep_fentry;
struct bpf_prog_desc nanosleep_fexit;
} progs;
struct {
int nanosleep_fentry_fd;
int nanosleep_fexit_fd;
} links;
struct fexit_sleep__bss {
int pid;
int fentry_cnt;
int fexit_cnt;
} *bss;
};
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210514003623.28033-18-alexei.starovoitov@gmail.com
The BPF program loading process performed by libbpf is quite complex
and consists of the following steps:
"open" phase:
- parse elf file and remember relocations, sections
- collect externs and ksyms including their btf_ids in prog's BTF
- patch BTF datasec (since llvm couldn't do it)
- init maps (old style map_def, BTF based, global data map, kconfig map)
- collect relocations against progs and maps
"load" phase:
- probe kernel features
- load vmlinux BTF
- resolve externs (kconfig and ksym)
- load program BTF
- init struct_ops
- create maps
- apply CO-RE relocations
- patch ld_imm64 insns with src_reg=PSEUDO_MAP, PSEUDO_MAP_VALUE, PSEUDO_BTF_ID
- reposition subprograms and adjust call insns
- sanitize and load progs
During this process libbpf does sys_bpf() calls to load BTF, create maps,
populate maps and finally load programs.
Instead of actually doing the syscalls generate a trace of what libbpf
would have done and represent it as the "loader program".
The "loader program" consists of single map with:
- union bpf_attr(s)
- BTF bytes
- map value bytes
- insns bytes
and single bpf program that passes bpf_attr(s) and data into bpf_sys_bpf() helper.
Executing such "loader program" via bpf_prog_test_run() command will
replay the sequence of syscalls that libbpf would have done which will result
the same maps created and programs loaded as specified in the elf file.
The "loader program" removes libelf and majority of libbpf dependency from
program loading process.
kconfig, typeless ksym, struct_ops and CO-RE are not supported yet.
The order of relocate_data and relocate_calls had to change, so that
bpf_gen__prog_load() can see all relocations for a given program with
correct insn_idx-es.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210514003623.28033-15-alexei.starovoitov@gmail.com
In order to be able to generate loader program in the later
patches change the order of data and text relocations.
Also improve the test to include data relos.
If the kernel supports "FD array" the map_fd relocations can be processed
before text relos since generated loader program won't need to manually
patch ld_imm64 insns with map_fd.
But ksym and kfunc relocations can only be processed after all calls
are relocated, since loader program will consist of a sequence
of calls to bpf_btf_find_by_name_kind() followed by patching of btf_id
and btf_obj_fd into corresponding ld_imm64 insns. The locations of those
ld_imm64 insns are specified in relocations.
Hence process all data relocations (maps, ksym, kfunc) together after call relos.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210514003623.28033-12-alexei.starovoitov@gmail.com
Add new helper:
long bpf_btf_find_by_name_kind(char *name, int name_sz, u32 kind, int flags)
Description
Find BTF type with given name and kind in vmlinux BTF or in module's BTFs.
Return
Returns btf_id and btf_obj_fd in lower and upper 32 bits.
It will be used by loader program to find btf_id to attach the program to
and to find btf_ids of ksyms.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210514003623.28033-10-alexei.starovoitov@gmail.com
Typical program loading sequence involves creating bpf maps and applying
map FDs into bpf instructions in various places in the bpf program.
This job is done by libbpf that is using compiler generated ELF relocations
to patch certain instruction after maps are created and BTFs are loaded.
The goal of fd_idx is to allow bpf instructions to stay immutable
after compilation. At load time the libbpf would still create maps as usual,
but it wouldn't need to patch instructions. It would store map_fds into
__u32 fd_array[] and would pass that pointer to sys_bpf(BPF_PROG_LOAD).
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210514003623.28033-9-alexei.starovoitov@gmail.com
With the addition of ssi_perf_data and ssi_perf_type struct signalfd_siginfo
is dangerously close to running out of space. All that remains is just
enough space for two additional 64bit fields. A practice of adding all
possible siginfo_t fields into struct singalfd_siginfo can not be supported
as adding the missing fields ssi_lower, ssi_upper, and ssi_pkey would
require two 64bit fields and one 32bit fields. In practice the fields
ssi_perf_data and ssi_perf_type can never be used by signalfd as the signal
that generates them always delivers them synchronously to the thread that
triggers them.
Therefore until someone actually needs the fields ssi_perf_data and
ssi_perf_type in signalfd_siginfo remove them. This leaves a bit more room
for future expansion.
v1: https://lkml.kernel.org/r/20210503203814.25487-12-ebiederm@xmission.com
v2: https://lkml.kernel.org/r/20210505141101.11519-12-ebiederm@xmission.com
Link: https://lkml.kernel.org/r/20210517195748.8880-5-ebiederm@xmission.com
Reviewed-by: Marco Elver <elver@google.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
