| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
A protection key fault is very similar to any other access error.
There must be a VMA, etc... We even want to take the same action
(SIGSEGV) that we do with a normal access fault.
However, we do need to let userspace know that something is
different. We do this the same way what we did with SEGV_BNDERR
with Memory Protection eXtensions (MPX): define a new SEGV code:
SEGV_PKUERR.
We add a siginfo field: si_pkey that reveals to userspace which
protection key was set on the PTE that we faulted on. There is
no other easy way for userspace to figure this out. They could
parse smaps but that would be a bit cruel.
We share space with in siginfo with _addr_bnd. #BR faults from
MPX are completely separate from page faults (#PF) that trigger
from protection key violations, so we never need both at the same
time.
Note that _pkey is a 64-bit value. The current hardware only
supports 4-bit protection keys. We do this because there is
_plenty_ of space in _sigfault and it is possible that future
processors would support more than 4 bits of protection keys.
The x86 code to actually fill in the siginfo is in the next
patch.
Signed-off-by: Dave Hansen <[email protected]>
Reviewed-by: Thomas Gleixner <[email protected]>
Cc: Al Viro <[email protected]>
Cc: Amanieu d'Antras <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Brian Gerst <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Oleg Nesterov <[email protected]>
Cc: Palmer Dabbelt <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Richard Weinberger <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Sasha Levin <[email protected]>
Cc: Vegard Nossum <[email protected]>
Cc: Vladimir Davydov <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
The init_name property of the device struct is sort of a hack and should
only be used for statically allocated devices. Since the device is
dynamically allocated here it is safe to use the proper way to set a
devices name by calling dev_set_name().
Signed-off-by: Lars-Peter Clausen <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
|
|
Remove the ftrace module notifier in favor of directly calling
ftrace_module_enable() and ftrace_release_mod() in the module loader.
Hard-coding the function calls directly in the module loader removes
dependence on the module notifier call chain and provides better
visibility and control over what gets called when, which is important
to kernel utilities such as livepatch.
This fixes a notifier ordering issue in which the ftrace module notifier
(and hence ftrace_module_enable()) for coming modules was being called
after klp_module_notify(), which caused livepatch modules to initialize
incorrectly. This patch removes dependence on the module notifier call
chain in favor of hard coding the corresponding function calls in the
module loader. This ensures that ftrace and livepatch code get called in
the correct order on patch module load and unload.
Fixes: 5156dca34a3e ("ftrace: Fix the race between ftrace and insmod")
Signed-off-by: Jessica Yu <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Reviewed-by: Petr Mladek <[email protected]>
Acked-by: Rusty Russell <[email protected]>
Reviewed-by: Josh Poimboeuf <[email protected]>
Reviewed-by: Miroslav Benes <[email protected]>
Signed-off-by: Jiri Kosina <[email protected]>
|
|
When dealing with key handling for shared futexes, we can drastically reduce
the usage/need of the page lock. 1) For anonymous pages, the associated futex
object is the mm_struct which does not require the page lock. 2) For inode
based, keys, we can check under RCU read lock if the page mapping is still
valid and take reference to the inode. This just leaves one rare race that
requires the page lock in the slow path when examining the swapcache.
Additionally realtime users currently have a problem with the page lock being
contended for unbounded periods of time during futex operations.
Task A
get_futex_key()
lock_page()
---> preempted
Now any other task trying to lock that page will have to wait until
task A gets scheduled back in, which is an unbound time.
With this patch, we pretty much have a lockless futex_get_key().
Experiments show that this patch can boost/speedup the hashing of shared
futexes with the perf futex benchmarks (which is good for measuring such
change) by up to 45% when there are high (> 100) thread counts on a 60 core
Westmere. Lower counts are pretty much in the noise range or less than 10%,
but mid range can be seen at over 30% overall throughput (hash ops/sec).
This makes anon-mem shared futexes much closer to its private counterpart.
Signed-off-by: Mel Gorman <[email protected]>
[ Ported on top of thp refcount rework, changelog, comments, fixes. ]
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Thomas Gleixner <[email protected]>
Cc: Chris Mason <[email protected]>
Cc: Darren Hart <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Sebastian Andrzej Siewior <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Ingo suggested we rename how we reference barriers A and B
regarding futex ordering guarantees. This patch replaces,
for both barriers, MB (A) with smp_mb(); (A), such that:
- We explicitly state that the barriers are SMP, and
- We standardize how we reference these across futex.c
helping readers follow what barrier does what and where.
Suggested-by: Ingo Molnar <[email protected]>
Signed-off-by: Davidlohr Bueso <[email protected]>
Reviewed-by: Thomas Gleixner <[email protected]>
Cc: Chris Mason <[email protected]>
Cc: Darren Hart <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Sebastian Andrzej Siewior <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
No functional change, just less confusing to read.
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Vince Weaver <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Signed-off-by: Ingo Molnar <[email protected]>
|
|
If CPU_UP_PREPARE is called it is not guaranteed, that a previously allocated
and assigned hash has been freed already, but perf_event_init_cpu()
unconditionally allocates and assignes a new hash if the swhash is referenced.
By overwriting the pointer the existing hash is not longer accessible.
Verify that there is no hash assigned on this cpu before allocating and
assigning a new one.
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Vince Weaver <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
If CPU_DOWN_PREPARE fails the perf hotplug notifier is called for
CPU_DOWN_FAILED and calls perf_event_init_cpu(), which checks whether the
swhash is referenced. If yes it allocates a new hash and stores the pointer in
the per cpu data structure.
But at this point the cpu is still online, so there must be a valid hash
already. By overwriting the pointer the existing hash is not longer
accessible.
Remove the CPU_DOWN_FAILED state, as there is nothing to (re)allocate.
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Vince Weaver <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
If CPU_UP_PREPARE fails the perf hotplug code calls perf_event_exit_cpu(),
which is a pointless exercise. The cpu is not online, so the smp function
calls return -ENXIO. So the result is a list walk to call noops.
Remove it.
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Vince Weaver <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
It's "too much" not "to much".
Signed-off-by: Steven Rostedt <[email protected]>
Acked-by: Juri Lelli <[email protected]>
Cc: Jiri Kosina <[email protected]>
Cc: Juri Lelli <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Remove an unnecessary assignment of variable not used any more.
( This has no runtime effects as GCC is smart enough to optimize
this out. )
Signed-off-by: Byungchul Park <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
[ Edited the changelog. ]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
allowing root in a non-init user namespace to mount it. This should
now be safe, because
1. non-init-root cannot mount a previously unbound subsystem
2. the task doing the mount must be privileged with respect to the
user namespace owning the cgroup namespace
3. the mounted subsystem will have its current cgroup as the root dentry.
the permissions will be unchanged, so tasks will receive no new
privilege over the cgroups which they did not have on the original
mounts.
Signed-off-by: Serge Hallyn <[email protected]>
|
|
This patch enables cgroup mounting inside userns when a process
as appropriate privileges. The cgroup filesystem mounted is
rooted at the cgroupns-root. Thus, in a container-setup, only
the hierarchy under the cgroupns-root is exposed inside the container.
This allows container management tools to run inside the containers
without depending on any global state.
Signed-off-by: Serge Hallyn <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
|
|
setns on a cgroup namespace is allowed only if
task has CAP_SYS_ADMIN in its current user-namespace and
over the user-namespace associated with target cgroupns.
No implicit cgroup changes happen with attaching to another
cgroupns. It is expected that the somone moves the attaching
process under the target cgroupns-root.
Signed-off-by: Aditya Kali <[email protected]>
Signed-off-by: Serge E. Hallyn <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
|
|
Introduce the ability to create new cgroup namespace. The newly created
cgroup namespace remembers the cgroup of the process at the point
of creation of the cgroup namespace (referred as cgroupns-root).
The main purpose of cgroup namespace is to virtualize the contents
of /proc/self/cgroup file. Processes inside a cgroup namespace
are only able to see paths relative to their namespace root
(unless they are moved outside of their cgroupns-root, at which point
they will see a relative path from their cgroupns-root).
For a correctly setup container this enables container-tools
(like libcontainer, lxc, lmctfy, etc.) to create completely virtualized
containers without leaking system level cgroup hierarchy to the task.
This patch only implements the 'unshare' part of the cgroupns.
Signed-off-by: Aditya Kali <[email protected]>
Signed-off-by: Serge Hallyn <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
|
|
For protection keys, we need to understand whether protections
should be enforced in software or not. In general, we enforce
protections when working on our own task, but not when on others.
We call these "current" and "remote" operations.
This patch introduces a new get_user_pages() variant:
get_user_pages_remote()
Which is a replacement for when get_user_pages() is called on
non-current tsk/mm.
We also introduce a new gup flag: FOLL_REMOTE which can be used
for the "__" gup variants to get this new behavior.
The uprobes is_trap_at_addr() location holds mmap_sem and
calls get_user_pages(current->mm) on an instruction address. This
makes it a pretty unique gup caller. Being an instruction access
and also really originating from the kernel (vs. the app), I opted
to consider this a 'remote' access where protection keys will not
be enforced.
Without protection keys, this patch should not change any behavior.
Signed-off-by: Dave Hansen <[email protected]>
Reviewed-by: Thomas Gleixner <[email protected]>
Cc: Andrea Arcangeli <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Brian Gerst <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: Kirill A. Shutemov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Naoya Horiguchi <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Srikar Dronamraju <[email protected]>
Cc: Vlastimil Babka <[email protected]>
Cc: [email protected]
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
This patch fix spelling typos found in printk and Kconfig.
Signed-off-by: Masanari Iida <[email protected]>
Acked-by: Randy Dunlap <[email protected]>
Signed-off-by: Jiri Kosina <[email protected]>
|
|
The irq code browses the list of actions differently to inspect the element
one by one. Even if it is not a problem, for the sake of consistent code,
provide a macro similar to for_each_irq_desc in order to have the same loop to
go through the actions list and use it in the code.
[ tglx: Renamed the macro ]
Signed-off-by: Daniel Lezcano <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
|
|
We want the fixes in here, and this resolves a merge error in tty_io.c
Signed-off-by: Greg Kroah-Hartman <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull lockdep fix from Thomas Gleixner:
"A single fix for the stack trace caching logic in lockdep, where the
duplicate avoidance managed to store no back trace at all"
* 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
locking/lockdep: Fix stack trace caching logic
|
|
It simplifies it and allows wide kick to be performed, even when IRQs
are disabled, without an asynchronous level in the middle.
This comes at a cost of some more overhead on features like perf and
posix cpu timers slow-paths, which is probably not much important
for nohz full users.
Requested-by: Peter Zijlstra <[email protected]>
Reviewed-by: Chris Metcalf <[email protected]>
Cc: Christoph Lameter <[email protected]>
Cc: Chris Metcalf <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Luiz Capitulino <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Viresh Kumar <[email protected]>
Signed-off-by: Frederic Weisbecker <[email protected]>
|
|
Export fetch_or() that's implemented and used internally by the
scheduler. We are going to use it for NO_HZ so make it generally
available.
Reviewed-by: Chris Metcalf <[email protected]>
Cc: Christoph Lameter <[email protected]>
Cc: Chris Metcalf <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Luiz Capitulino <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Viresh Kumar <[email protected]>
Signed-off-by: Frederic Weisbecker <[email protected]>
|
|
Conflicts:
kernel/locking/lockdep.c
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Testing cgroup2 can be painful with system software automatically
mounting and populating all cgroup controllers in v1 mode. Sometimes
they can be unmounted from rc.local, sometimes even that is too late.
Provide a commandline option to disable certain controllers in v1
mounts, so that they remain available for cgroup2 mounts.
Example use:
cgroup_no_v1=memory,cpu
cgroup_no_v1=all
Disabling will be confirmed at boot-time as such:
[ 0.013770] Disabling cpu control group subsystem in v1 mounts
[ 0.016004] Disabling memory control group subsystem in v1 mounts
Signed-off-by: Johannes Weiner <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
|
|
The pfn_t type uses an unsigned long to store a pfn + flags value. On a
64-bit platform the upper 12 bits of an unsigned long are never used for
storing the value of a pfn. However, this is not true on highmem
platforms, all 32-bits of a pfn value are used to address a 44-bit
physical address space. A pfn_t needs to store a 64-bit value.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=112211
Fixes: 01c8f1c44b83 ("mm, dax, gpu: convert vm_insert_mixed to pfn_t")
Signed-off-by: Dan Williams <[email protected]>
Reported-by: Stuart Foster <[email protected]>
Reported-by: Julian Margetson <[email protected]>
Tested-by: Julian Margetson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Mike said:
: CONFIG_UBSAN_ALIGNMENT breaks x86-64 kernel with lockdep enabled, i. e
: kernel with CONFIG_UBSAN_ALIGNMENT fails to load without even any error
: message.
:
: The problem is that ubsan callbacks use spinlocks and might be called
: before lockdep is initialized. Particularly this line in the
: reserve_ebda_region function causes problem:
:
: lowmem = *(unsigned short *)__va(BIOS_LOWMEM_KILOBYTES);
:
: If i put lockdep_init() before reserve_ebda_region call in
: x86_64_start_reservations kernel loads well.
Fix this ordering issue permanently: change lockdep so that it uses
hlists for the hash tables. Unlike a list_head, an hlist_head is in its
initialized state when it is all-zeroes, so lockdep is ready for
operation immediately upon boot - lockdep_init() need not have run.
The patch will also save some memory.
lockdep_init() and lockdep_initialized can be done away with now - a 4.6
patch has been prepared to do this.
Reported-by: Mike Krinkin <[email protected]>
Suggested-by: Mike Krinkin <[email protected]>
Cc: Andrey Ryabinin <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Pull networking fixes from David Miller:
1) Fix BPF handling of branch offset adjustmnets on backjumps, from
Daniel Borkmann.
2) Make sure selinux knows about SOCK_DESTROY netlink messages, from
Lorenzo Colitti.
3) Fix openvswitch tunnel mtu regression, from David Wragg.
4) Fix ICMP handling of TCP sockets in syn_recv state, from Eric
Dumazet.
5) Fix SCTP user hmacid byte ordering bug, from Xin Long.
6) Fix recursive locking in ipv6 addrconf, from Subash Abhinov
Kasiviswanathan.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
bpf: fix branch offset adjustment on backjumps after patching ctx expansion
vxlan, gre, geneve: Set a large MTU on ovs-created tunnel devices
geneve: Relax MTU constraints
vxlan: Relax MTU constraints
flow_dissector: Fix unaligned access in __skb_flow_dissector when used by eth_get_headlen
of: of_mdio: Add marvell, 88e1145 to whitelist of PHY compatibilities.
selinux: nlmsgtab: add SOCK_DESTROY to the netlink mapping tables
sctp: translate network order to host order when users get a hmacid
enic: increment devcmd2 result ring in case of timeout
tg3: Fix for tg3 transmit queue 0 timed out when too many gso_segs
net:Add sysctl_max_skb_frags
tcp: do not drop syn_recv on all icmp reports
ipv6: fix a lockdep splat
unix: correctly track in-flight fds in sending process user_struct
update be2net maintainers' email addresses
dwc_eth_qos: Reset hardware before PHY start
ipv6: addrconf: Fix recursive spin lock call
|
|
replacing printk(s) with appropriate pr_info and pr_err
in order to fix checkpatch.pl warnings
Signed-off-by: Saurabh Sengar <[email protected]>
Acked-by: Pavel Machek <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
Wall time obtained from do_gettimeofday gives 32 bit timeval which can only
represent time until January 2038. This patch moves to ktime_t, a 64-bit time.
Also, wall time is susceptible to sudden jumps due to user setting the time or
due to NTP. Boot time is constantly increasing time better suited for
subtracting two timestamps.
Signed-off-by: Abhilash Jindal <[email protected]>
Acked-by: Pavel Machek <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
|
|
|
|
When ctx access is used, the kernel often needs to expand/rewrite
instructions, so after that patching, branch offsets have to be
adjusted for both forward and backward jumps in the new eBPF program,
but for backward jumps it fails to account the delta. Meaning, for
example, if the expansion happens exactly on the insn that sits at
the jump target, it doesn't fix up the back jump offset.
Analysis on what the check in adjust_branches() is currently doing:
/* adjust offset of jmps if necessary */
if (i < pos && i + insn->off + 1 > pos)
insn->off += delta;
else if (i > pos && i + insn->off + 1 < pos)
insn->off -= delta;
First condition (forward jumps):
Before: After:
insns[0] insns[0]
insns[1] <--- i/insn insns[1] <--- i/insn
insns[2] <--- pos insns[P] <--- pos
insns[3] insns[P] `------| delta
insns[4] <--- target_X insns[P] `-----|
insns[5] insns[3]
insns[4] <--- target_X
insns[5]
First case is if we cross pos-boundary and the jump instruction was
before pos. This is handeled correctly. I.e. if i == pos, then this
would mean our jump that we currently check was the patchlet itself
that we just injected. Since such patchlets are self-contained and
have no awareness of any insns before or after the patched one, the
delta is correctly not adjusted. Also, for the second condition in
case of i + insn->off + 1 == pos, means we jump to that newly patched
instruction, so no offset adjustment are needed. That part is correct.
Second condition (backward jumps):
Before: After:
insns[0] insns[0]
insns[1] <--- target_X insns[1] <--- target_X
insns[2] <--- pos <-- target_Y insns[P] <--- pos <-- target_Y
insns[3] insns[P] `------| delta
insns[4] <--- i/insn insns[P] `-----|
insns[5] insns[3]
insns[4] <--- i/insn
insns[5]
Second interesting case is where we cross pos-boundary and the jump
instruction was after pos. Backward jump with i == pos would be
impossible and pose a bug somewhere in the patchlet, so the first
condition checking i > pos is okay only by itself. However, i +
insn->off + 1 < pos does not always work as intended to trigger the
adjustment. It works when jump targets would be far off where the
delta wouldn't matter. But, for example, where the fixed insn->off
before pointed to pos (target_Y), it now points to pos + delta, so
that additional room needs to be taken into account for the check.
This means that i) both tests here need to be adjusted into pos + delta,
and ii) for the second condition, the test needs to be <= as pos
itself can be a target in the backjump, too.
Fixes: 9bac3d6d548e ("bpf: allow extended BPF programs access skb fields")
Signed-off-by: Daniel Borkmann <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
Pull cgroup fixes from Tejun Heo:
- The destruction path of cgroup objects are asynchronous and
multi-staged and some of them ended up destroying parents before
children leading to failures in cpu and memory controllers. Ensure
that parents are always destroyed after children.
- cpuset mm node migration was performed synchronously while holding
threadgroup and cgroup mutexes and the recent threadgroup locking
update resulted in a possible deadlock. The migration is best effort
and shouldn't have been performed under those locks to begin with.
Made asynchronous.
- Minor documentation fix.
* 'for-4.5-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
Documentation: cgroup: Fix 'cgroup-legacy' -> 'cgroup-v1'
cgroup: make sure a parent css isn't freed before its children
cgroup: make sure a parent css isn't offlined before its children
cpuset: make mm migration asynchronous
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq
Pull workqueue fixes from Tejun Heo:
"Workqueue fixes for v4.5-rc3.
- Remove a spurious triggering of flush dependency warning.
- Officially break local execution guarantee of unbound work items
and add a debug feature to flush out usages which depend on it.
- Work around CPU -> NODE mapping becoming invalid on CPU offline.
The branch is young but pushing out early as stable kernels are being
affected"
* 'for-4.5-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
workqueue: handle NUMA_NO_NODE for unbound pool_workqueue lookup
workqueue: implement "workqueue.debug_force_rr_cpu" debug feature
workqueue: schedule WORK_CPU_UNBOUND work on wq_unbound_cpumask CPUs
Revert "workqueue: make sure delayed work run in local cpu"
workqueue: skip flush dependency checks for legacy workqueues
|
|
When looking up the pool_workqueue to use for an unbound workqueue,
workqueue assumes that the target CPU is always bound to a valid NUMA
node. However, currently, when a CPU goes offline, the mapping is
destroyed and cpu_to_node() returns NUMA_NO_NODE.
This has always been broken but hasn't triggered often enough before
874bbfe600a6 ("workqueue: make sure delayed work run in local cpu").
After the commit, workqueue forcifully assigns the local CPU for
delayed work items without explicit target CPU to fix a different
issue. This widens the window where CPU can go offline while a
delayed work item is pending causing delayed work items dispatched
with target CPU set to an already offlined CPU. The resulting
NUMA_NO_NODE mapping makes workqueue try to queue the work item on a
NULL pool_workqueue and thus crash.
While 874bbfe600a6 has been reverted for a different reason making the
bug less visible again, it can still happen. Fix it by mapping
NUMA_NO_NODE to the default pool_workqueue from unbound_pwq_by_node().
This is a temporary workaround. The long term solution is keeping CPU
-> NODE mapping stable across CPU off/online cycles which is being
worked on.
Signed-off-by: Tejun Heo <[email protected]>
Reported-by: Mike Galbraith <[email protected]>
Cc: Tang Chen <[email protected]>
Cc: Rafael J. Wysocki <[email protected]>
Cc: Len Brown <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/g/[email protected]
Link: http://lkml.kernel.org/g/[email protected]
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux
Pull module fixes from Rusty Russell:
"Fix for async_probe module param added in 4.3 (clearly not widely used
yet), and a much more interesting kallsyms race which has been around
approximately forever. This fix is more invasive, and will require
some care in backporting, but I hated all the bandaids I could think
of, so...
There are some more coming, which are only for breakages introduced
this cycle (livepatch), but wanted these in now"
* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux:
modules: fix longstanding /proc/kallsyms vs module insertion race.
module: wrapper for symbol name.
modules: fix modparam async_probe request
|
|
Workqueue used to guarantee local execution for work items queued
without explicit target CPU. The guarantee is gone now which can
break some usages in subtle ways. To flush out those cases, this
patch implements a debug feature which forces round-robin CPU
selection for all such work items.
The debug feature defaults to off and can be enabled with a kernel
parameter. The default can be flipped with a debug config option.
If you hit this commit during bisection, please refer to 041bd12e272c
("Revert "workqueue: make sure delayed work run in local cpu"") for
more information and ping me.
Signed-off-by: Tejun Heo <[email protected]>
|
|
WORK_CPU_UNBOUND work items queued to a bound workqueue always run
locally. This is a good thing normally, but not when the user has
asked us to keep unbound work away from certain CPUs. Round robin
these to wq_unbound_cpumask CPUs instead, as perturbation avoidance
trumps performance.
tj: Cosmetic and comment changes. WARN_ON_ONCE() dropped from empty
(wq_unbound_cpumask AND cpu_online_mask). If we want that, it
should be done when config changes.
Signed-off-by: Mike Galbraith <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
|
|
This reverts commit 874bbfe600a660cba9c776b3957b1ce393151b76.
Workqueue used to implicity guarantee that work items queued without
explicit CPU specified are put on the local CPU. Recent changes in
timer broke the guarantee and led to vmstat breakage which was fixed
by 176bed1de5bf ("vmstat: explicitly schedule per-cpu work on the CPU
we need it to run on").
vmstat is the most likely to expose the issue and it's quite possible
that there are other similar problems which are a lot more difficult
to trigger. As a preventive measure, 874bbfe600a6 ("workqueue: make
sure delayed work run in local cpu") was applied to restore the local
CPU guarnatee. Unfortunately, the change exposed a bug in timer code
which got fixed by 22b886dd1018 ("timers: Use proper base migration in
add_timer_on()"). Due to code restructuring, the commit couldn't be
backported beyond certain point and stable kernels which only had
874bbfe600a6 started crashing.
The local CPU guarantee was accidental more than anything else and we
want to get rid of it anyway. As, with the vmstat case fixed,
874bbfe600a6 is causing more problems than it's fixing, it has been
decided to take the chance and officially break the guarantee by
reverting the commit. A debug feature will be added to force foreign
CPU assignment to expose cases relying on the guarantee and fixes for
the individual cases will be backported to stable as necessary.
Signed-off-by: Tejun Heo <[email protected]>
Fixes: 874bbfe600a6 ("workqueue: make sure delayed work run in local cpu")
Link: http://lkml.kernel.org/g/[email protected]
Cc: [email protected]
Cc: Mike Galbraith <[email protected]>
Cc: Henrique de Moraes Holschuh <[email protected]>
Cc: Daniel Bilik <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: Shaohua Li <[email protected]>
Cc: Sasha Levin <[email protected]>
Cc: Ben Hutchings <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Daniel Bilik <[email protected]>
Cc: Jiri Slaby <[email protected]>
Cc: Michal Hocko <[email protected]>
|
|
The pseudo-interleaving in NUMA placement has a fundamental problem:
using hard usage thresholds to spread memory equally between nodes
can prevent workloads from converging, or keep memory "trapped" on
nodes where the workload is barely running any more.
In order for workloads to properly converge, the memory migration
should not be stopped when nodes reach parity, but instead be
distributed according to how heavily memory is used from each node.
This way memory migration and task migration reinforce each other,
instead of one putting the brakes on the other.
Remove the hard thresholds from the pseudo-interleaving code, and
instead use a more gradual policy on memory placement. This also
seems to improve convergence of workloads that do not run flat out,
but sleep in between bursts of activity.
We still want to slow down NUMA scanning and migration once a workload
has settled on a few actively used nodes, so keep the 3/4 hysteresis
in place. Keep track of whether a workload is actively running on
multiple nodes, so task_numa_migrate does a full scan of the system
for better task placement.
In the case of running 3 SPECjbb2005 instances on a 4 node system,
this code seems to result in fairer distribution of memory between
nodes, with more memory bandwidth for each instance.
Signed-off-by: Rik van Riel <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
[ Minor readability tweaks. ]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Lockdep is initialized at compile time now. Get rid of lockdep_init().
Signed-off-by: Andrey Ryabinin <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Mike Krinkin <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Mike said:
: CONFIG_UBSAN_ALIGNMENT breaks x86-64 kernel with lockdep enabled, i.e.
: kernel with CONFIG_UBSAN_ALIGNMENT=y fails to load without even any error
: message.
:
: The problem is that ubsan callbacks use spinlocks and might be called
: before lockdep is initialized. Particularly this line in the
: reserve_ebda_region function causes problem:
:
: lowmem = *(unsigned short *)__va(BIOS_LOWMEM_KILOBYTES);
:
: If i put lockdep_init() before reserve_ebda_region call in
: x86_64_start_reservations kernel loads well.
Fix this ordering issue permanently: change lockdep so that it uses hlists
for the hash tables. Unlike a list_head, an hlist_head is in its
initialized state when it is all-zeroes, so lockdep is ready for operation
immediately upon boot - lockdep_init() need not have run.
The patch will also save some memory.
Probably lockdep_init() and lockdep_initialized can be done away with now.
Suggested-by: Mike Krinkin <[email protected]>
Reported-by: Mike Krinkin <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Cc: Andrey Ryabinin <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
schedstats is very useful during debugging and performance tuning but it
incurs overhead to calculate the stats. As such, even though it can be
disabled at build time, it is often enabled as the information is useful.
This patch adds a kernel command-line and sysctl tunable to enable or
disable schedstats on demand (when it's built in). It is disabled
by default as someone who knows they need it can also learn to enable
it when necessary.
The benefits are dependent on how scheduler-intensive the workload is.
If it is then the patch reduces the number of cycles spent calculating
the stats with a small benefit from reducing the cache footprint of the
scheduler.
These measurements were taken from a 48-core 2-socket
machine with Xeon(R) E5-2670 v3 cpus although they were also tested on a
single socket machine 8-core machine with Intel i7-3770 processors.
netperf-tcp
4.5.0-rc1 4.5.0-rc1
vanilla nostats-v3r1
Hmean 64 560.45 ( 0.00%) 575.98 ( 2.77%)
Hmean 128 766.66 ( 0.00%) 795.79 ( 3.80%)
Hmean 256 950.51 ( 0.00%) 981.50 ( 3.26%)
Hmean 1024 1433.25 ( 0.00%) 1466.51 ( 2.32%)
Hmean 2048 2810.54 ( 0.00%) 2879.75 ( 2.46%)
Hmean 3312 4618.18 ( 0.00%) 4682.09 ( 1.38%)
Hmean 4096 5306.42 ( 0.00%) 5346.39 ( 0.75%)
Hmean 8192 10581.44 ( 0.00%) 10698.15 ( 1.10%)
Hmean 16384 18857.70 ( 0.00%) 18937.61 ( 0.42%)
Small gains here, UDP_STREAM showed nothing intresting and neither did
the TCP_RR tests. The gains on the 8-core machine were very similar.
tbench4
4.5.0-rc1 4.5.0-rc1
vanilla nostats-v3r1
Hmean mb/sec-1 500.85 ( 0.00%) 522.43 ( 4.31%)
Hmean mb/sec-2 984.66 ( 0.00%) 1018.19 ( 3.41%)
Hmean mb/sec-4 1827.91 ( 0.00%) 1847.78 ( 1.09%)
Hmean mb/sec-8 3561.36 ( 0.00%) 3611.28 ( 1.40%)
Hmean mb/sec-16 5824.52 ( 0.00%) 5929.03 ( 1.79%)
Hmean mb/sec-32 10943.10 ( 0.00%) 10802.83 ( -1.28%)
Hmean mb/sec-64 15950.81 ( 0.00%) 16211.31 ( 1.63%)
Hmean mb/sec-128 15302.17 ( 0.00%) 15445.11 ( 0.93%)
Hmean mb/sec-256 14866.18 ( 0.00%) 15088.73 ( 1.50%)
Hmean mb/sec-512 15223.31 ( 0.00%) 15373.69 ( 0.99%)
Hmean mb/sec-1024 14574.25 ( 0.00%) 14598.02 ( 0.16%)
Hmean mb/sec-2048 13569.02 ( 0.00%) 13733.86 ( 1.21%)
Hmean mb/sec-3072 12865.98 ( 0.00%) 13209.23 ( 2.67%)
Small gains of 2-4% at low thread counts and otherwise flat. The
gains on the 8-core machine were slightly different
tbench4 on 8-core i7-3770 single socket machine
Hmean mb/sec-1 442.59 ( 0.00%) 448.73 ( 1.39%)
Hmean mb/sec-2 796.68 ( 0.00%) 794.39 ( -0.29%)
Hmean mb/sec-4 1322.52 ( 0.00%) 1343.66 ( 1.60%)
Hmean mb/sec-8 2611.65 ( 0.00%) 2694.86 ( 3.19%)
Hmean mb/sec-16 2537.07 ( 0.00%) 2609.34 ( 2.85%)
Hmean mb/sec-32 2506.02 ( 0.00%) 2578.18 ( 2.88%)
Hmean mb/sec-64 2511.06 ( 0.00%) 2569.16 ( 2.31%)
Hmean mb/sec-128 2313.38 ( 0.00%) 2395.50 ( 3.55%)
Hmean mb/sec-256 2110.04 ( 0.00%) 2177.45 ( 3.19%)
Hmean mb/sec-512 2072.51 ( 0.00%) 2053.97 ( -0.89%)
In constract, this shows a relatively steady 2-3% gain at higher thread
counts. Due to the nature of the patch and the type of workload, it's
not a surprise that the result will depend on the CPU used.
hackbench-pipes
4.5.0-rc1 4.5.0-rc1
vanilla nostats-v3r1
Amean 1 0.0637 ( 0.00%) 0.0660 ( -3.59%)
Amean 4 0.1229 ( 0.00%) 0.1181 ( 3.84%)
Amean 7 0.1921 ( 0.00%) 0.1911 ( 0.52%)
Amean 12 0.3117 ( 0.00%) 0.2923 ( 6.23%)
Amean 21 0.4050 ( 0.00%) 0.3899 ( 3.74%)
Amean 30 0.4586 ( 0.00%) 0.4433 ( 3.33%)
Amean 48 0.5910 ( 0.00%) 0.5694 ( 3.65%)
Amean 79 0.8663 ( 0.00%) 0.8626 ( 0.43%)
Amean 110 1.1543 ( 0.00%) 1.1517 ( 0.22%)
Amean 141 1.4457 ( 0.00%) 1.4290 ( 1.16%)
Amean 172 1.7090 ( 0.00%) 1.6924 ( 0.97%)
Amean 192 1.9126 ( 0.00%) 1.9089 ( 0.19%)
Some small gains and losses and while the variance data is not included,
it's close to the noise. The UMA machine did not show anything particularly
different
pipetest
4.5.0-rc1 4.5.0-rc1
vanilla nostats-v2r2
Min Time 4.13 ( 0.00%) 3.99 ( 3.39%)
1st-qrtle Time 4.38 ( 0.00%) 4.27 ( 2.51%)
2nd-qrtle Time 4.46 ( 0.00%) 4.39 ( 1.57%)
3rd-qrtle Time 4.56 ( 0.00%) 4.51 ( 1.10%)
Max-90% Time 4.67 ( 0.00%) 4.60 ( 1.50%)
Max-93% Time 4.71 ( 0.00%) 4.65 ( 1.27%)
Max-95% Time 4.74 ( 0.00%) 4.71 ( 0.63%)
Max-99% Time 4.88 ( 0.00%) 4.79 ( 1.84%)
Max Time 4.93 ( 0.00%) 4.83 ( 2.03%)
Mean Time 4.48 ( 0.00%) 4.39 ( 1.91%)
Best99%Mean Time 4.47 ( 0.00%) 4.39 ( 1.91%)
Best95%Mean Time 4.46 ( 0.00%) 4.38 ( 1.93%)
Best90%Mean Time 4.45 ( 0.00%) 4.36 ( 1.98%)
Best50%Mean Time 4.36 ( 0.00%) 4.25 ( 2.49%)
Best10%Mean Time 4.23 ( 0.00%) 4.10 ( 3.13%)
Best5%Mean Time 4.19 ( 0.00%) 4.06 ( 3.20%)
Best1%Mean Time 4.13 ( 0.00%) 4.00 ( 3.39%)
Small improvement and similar gains were seen on the UMA machine.
The gain is small but it stands to reason that doing less work in the
scheduler is a good thing. The downside is that the lack of schedstats and
tracepoints may be surprising to experts doing performance analysis until
they find the existence of the schedstats= parameter or schedstats sysctl.
It will be automatically activated for latencytop and sleep profiling to
alleviate the problem. For tracepoints, there is a simple warning as it's
not safe to activate schedstats in the context when it's known the tracepoint
may be wanted but is unavailable.
Signed-off-by: Mel Gorman <[email protected]>
Reviewed-by: Matt Fleming <[email protected]>
Reviewed-by: Srikar Dronamraju <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
When doing ebpf+kprobe on some hot TCP functions (e.g.
tcp_rcv_established), the kprobe_dispatcher() function
shows up in 'perf report'.
In kprobe_dispatcher(), there is a lot of cache bouncing
on 'tk->nhit++'. 'tk->nhit' and 'tk->tp.flags' also share
the same cacheline.
perf report (cycles:pp):
8.30% ipv4_dst_check
4.74% copy_user_enhanced_fast_string
3.93% dst_release
2.80% tcp_v4_rcv
2.31% queued_spin_lock_slowpath
2.30% _raw_spin_lock
1.88% mlx4_en_process_rx_cq
1.84% eth_get_headlen
1.81% ip_rcv_finish
~~~~
1.71% kprobe_dispatcher
~~~~
1.55% mlx4_en_xmit
1.09% __probe_kernel_read
perf report after patch:
9.15% ipv4_dst_check
5.00% copy_user_enhanced_fast_string
4.12% dst_release
2.96% tcp_v4_rcv
2.50% _raw_spin_lock
2.39% queued_spin_lock_slowpath
2.11% eth_get_headlen
2.03% mlx4_en_process_rx_cq
1.69% mlx4_en_xmit
1.19% ip_rcv_finish
1.12% __probe_kernel_read
1.02% ehci_hcd_cleanup
Signed-off-by: Martin KaFai Lau <[email protected]>
Acked-by: Masami Hiramatsu <[email protected]>
Acked-by: Steven Rostedt <[email protected]>
Cc: Alexei Starovoitov <[email protected]>
Cc: Josef Bacik <[email protected]>
Cc: Kernel Team <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
check_prev_add() caches saved stack trace in static trace variable
to avoid duplicate save_trace() calls in dependencies involving trylocks.
But that caching logic contains a bug. We may not save trace on first
iteration due to early return from check_prev_add(). Then on the
second iteration when we actually need the trace we don't save it
because we think that we've already saved it.
Let check_prev_add() itself control when stack is saved.
There is another bug. Trace variable is protected by graph lock.
But we can temporary release graph lock during printing.
Fix this by invalidating cached stack trace when we release graph lock.
Signed-off-by: Dmitry Vyukov <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Signed-off-by: Weiyuan <[email protected]>
Signed-off-by: Paul Moore <[email protected]>
|
|
If we isolate CPUs, then we don't want random device interrupts on them. Even
w/o the user space irq balancer enabled we can end up with irqs on non boot
cpus and chasing newly requested interrupts is a tedious task.
Allow to restrict the default irq affinity mask.
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Frederic Weisbecker <[email protected]>
Cc: Chris Metcalf <[email protected]>
Cc: Christoph Lameter <[email protected]>
Cc: Sebastian Siewior <[email protected]>
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1602031948190.25254@nanos
Signed-off-by: Thomas Gleixner <[email protected]>
|
|
The functions bpf_map_lookup_elem(map, key, value) and
bpf_map_update_elem(map, key, value, flags) need to get/set
values from all-cpus for per-cpu hash and array maps,
so that user space can aggregate/update them as necessary.
Example of single counter aggregation in user space:
unsigned int nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
long values[nr_cpus];
long value = 0;
bpf_lookup_elem(fd, key, values);
for (i = 0; i < nr_cpus; i++)
value += values[i];
The user space must provide round_up(value_size, 8) * nr_cpus
array to get/set values, since kernel will use 'long' copy
of per-cpu values to try to copy good counters atomically.
It's a best-effort, since bpf programs and user space are racing
to access the same memory.
Signed-off-by: Alexei Starovoitov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Primary use case is a histogram array of latency
where bpf program computes the latency of block requests or other
events and stores histogram of latency into array of 64 elements.
All cpus are constantly running, so normal increment is not accurate,
bpf_xadd causes cache ping-pong and this per-cpu approach allows
fastest collision-free counters.
Signed-off-by: Alexei Starovoitov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Introduce BPF_MAP_TYPE_PERCPU_HASH map type which is used to do
accurate counters without need to use BPF_XADD instruction which turned
out to be too costly for high-performance network monitoring.
In the typical use case the 'key' is the flow tuple or other long
living object that sees a lot of events per second.
bpf_map_lookup_elem() returns per-cpu area.
Example:
struct {
u32 packets;
u32 bytes;
} * ptr = bpf_map_lookup_elem(&map, &key);
/* ptr points to this_cpu area of the value, so the following
* increments will not collide with other cpus
*/
ptr->packets ++;
ptr->bytes += skb->len;
bpf_update_elem() atomically creates a new element where all per-cpu
values are zero initialized and this_cpu value is populated with
given 'value'.
Note that non-per-cpu hash map always allocates new element
and then deletes old after rcu grace period to maintain atomicity
of update. Per-cpu hash map updates element values in-place.
Signed-off-by: Alexei Starovoitov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|