18 Dec, 2012
40 commits
-
With this change, the aoe driver treats the value zero as special for
the aoe_deadsecs module parameter. Normally, this value specifies the
number of seconds during which the driver will continue to attempt
retransmits to an unresponsive AoE target. After aoe_deadsecs has
elapsed, the aoe driver marks the aoe device as "down" and fails all
I/O.The new meaning of an aoe_deadsecs of zero is for the driver to
retransmit commands indefinitely.Signed-off-by: Ed Cashin
Cc: Jens Axboe
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Many AoE targets have four or fewer network ports, but some existing
storage devices have many, and the AoE protocol sets no limit.This patch allows the use of more than eight remote MAC addresses per AoE
target, while reducing the amount of memory used by the aoe driver in
cases where there are many AoE targets with fewer than eight MAC addresses
each.Signed-off-by: Ed Cashin
Cc: Jens Axboe
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
This change avoids a race that could result in a NULL pointer derference
following a WARNing from kobject_add_internal, "don't try to register
things with the same name in the same directory."The problem was found with a test that forgets and discovers an
aoe device in a loop:while test ! -r /tmp/stop; do
aoe-flush -a
aoe-discover
doneThe race was between aoedev_flush taking aoedevs out of the devlist,
allowing a new discovery of the same AoE target to take place before the
driver gets around to calling sysfs_remove_group. Fixing that one
revealed another race between do_open and add_disk, and this patch avoids
that, too.The fix required some care, because for flushing (forgetting) an aoedev,
some of the steps must be performed under lock and some must be able to
sleep. Also, for discovering a new aoedev, some steps might sleep.The check for a bad aoedev pointer remains from a time when about half of
this patch was done, and it was possible for the
bdev->bd_disk->private_data to become corrupted. The check should be
removed eventually, but it is not expected to add significant overhead,
occurring in the aoeblk_open routine.Signed-off-by: Ed Cashin
Cc: Jens Axboe
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
An AoE target can have multiple network ports used for AoE, and in the
aoe driver, those are tracked by the aoetgt struct. These changes allow
the aoe driver to handle network paths, or aoetgts, that are not working
well, compared to the others.Paths that do not get responses despite the retransmission of AoE
commands are marked as "tainted", and non-tainted paths are preferred.Meanwhile, the aoe driver attempts to "probe" the tainted path in the
background by issuing reads of LBA 0 that are padded out to full
(possibly jumbo-frame) size. If the probes get responses, then the path
is "redeemed", and its taint is removed.This mechanism has been shown to be helpful in transparently handling
and recovering from real-world network "brown outs" in ways that the
earlier "shoot the help-needing target in the head" mechanism could not.Signed-off-by: Ed Cashin
Cc: Jens Axboe
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The value returned by the static minor device number number allocator is
the real minor number, so it must be multiplied by the supported number
of partitions per aoedev.Without this fix the support for systems without udev is incomplete, and
the few users of aoe on such systems will have surprising results when
device nodes names do not match the AoE target.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Because the minor_get and related functions use the return values for
errors, the compiler doesn't know that sysminor will always either 1) be
initialized in aoedev_by_aoeaddr by the call to minor_get, or 2) be
unused as the "goto out" is executed.This patch avoids the compiler warning.
Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
For some special-purpose systems where udev isn't present, static
allocation of minor numbers is desirable. This update distinguishes
different failure scenarios, to help the user understand what went
wrong.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
There is no need to call the request handler function in the I/O
completion routine. The user impact of not doing it is a more "nice" aoe
driver that is less susceptible to causing soft lockups.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
A misplaced comment was attached to the nout member of the aoetgt. This
change corrects the comment.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The aoe driver will never be waiting for more than aoe_maxout AoE
commands from a given remote network port on an AoE target. Increasing
the cap increases performance. Users can tighten the setting to reduce
the amount of memory used for handling AoE traffic or the network
bandwidth used for AoE.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Before the aoe driver was an I/O request handler, it was a
make_request-style block driver. Even so, there was a problem where
sysfs expected a request queue to exist, so one was provided in commit
7135a71b19be ("aoe: allocate unused request_queue for sysfs").During the transition to the request-handler style, a patch was merged
that was based on a driver without the noop queue, and the noop queue
remained in place after the patch was merged, even though a new
functional queue was introduced by the patch, allocated through
blk_init_queue.The user impact is a memory leak proportional to the number of AoE
targets discovered. This patch removes the memory leak and cleans up
vestiges of the old do-nothing queue from the aoeblk_gdalloc function.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Commit f3b8e07af774 ("aoe: commands in retransmit queue use new
destination on failure") omits the copying of the coarse-grained time
when an AoE command was sent during the failover from one destination
MAC address on the AoE target to another.The coarse-grained timing is only used when the system time changes or
an unlikely length of time has passed since the sending of the AoE
command. Users will not be impacted unless their system clock is very
inaccurate or something unusual (e.g., 10 GbE link reset) happens during
the period when the aoe driver is handling the failure of a port on the
AoE target. Being effected will mean that an AoE target could be
considered "down" too eagerly.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
When one remote MAC address isn't working as a destination for AoE
commands, the frames used to track information associated with the AoE
commands are moved to a new aoetgt (defined by the tuple of {AoE major,
AoE minor, target MAC address}).This patch makes sure that the frames on the queue for retransmits that
need to be done are updated to use the new destination, so that
retransmits will be sent through a working network path.Without this change, packets on the retransmit queue will be needlessly
retransmitted to the unresponsive destination MAC, possibly causing
premature target failure before there's time for the retransmit timer to
run again, decide to retransmit again, and finally update the destination
to a working MAC address on the AoE target.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
These changes improve the accuracy of the decision about whether it's time
to retransmit an AoE command by using the microsecond-resolution
gettimeofday instead of jiffies.Because the system time can jump suddenly, the decision reverts to using
jiffies if the high-resolution time difference is relatively large.
Otherwise the AoE targets could be considered failed inappropriately.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
With this bugfix in place the calculation of the criterion for "lateness"
is performed under lock. Without the lock, there is a chance that one of
the non-atomic operations performed on the round trip time statistics
could be incomplete, such that an incorrect lateness criterion would be
calculated.Without this change, the effect of the bug would be rare unecessary but
benign retransmissions.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The /dev/etherd/err character device provides low-level information about
normal but sometimes interesting AoE command retransmits and "unexpected
responses", i.e., responses for packets that have already been
retransmitted.This change adds MAC addresses to the messages about unexpected responses,
so that when they occur, it's more easy to determine the network paths to
which they belong.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The aoe driver already had some congestion handling, but it was limited in
its ability to cope with the kind of congestion that can arise on more
complex networks such as those involving paths through multiple ethernet
switches.Some of the lessons from TCP's history of development can be applied to
improving the congestion control and avoidance on AoE storage networks.
These changes use familar concepts from Van Jacobson's "Congestion
Avoidance and Control" paper from '88, without adding significant
overhead.This patch depends on an upcoming patch that covers the failover case when
AoE commands being retransmitted are transferred from one retransmit queue
to another. Another upcoming patch increases the timing accuracy.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Make the aoe driver follow expected behavior when the user uses ioctl to
get the ATA device identify information, allowing access to model, serial
number, etc.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The userland aoetools package includes an "aoe-stat" command that can
display a "payload size" column when the aoe driver exports this
information. Users can quickly see what amount of user data is
transferred inside each AoE command on the network, network headers
excluded.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The GPFS filesystem is an example of an aoe user that requires the aoe
driver to support I/O request sizes larger than the default. Most users
will not need large I/O request sizes, because they would need to be split
up into multiple AoE commands anyway.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Users sometimes want to cause the aoe driver to forget a particular
previously discovered device when it is no longer online. The aoetools
provide an "aoe-flush" command that users run to perform this
administrative task. The changes below provide the support needed in the
driver.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The ATA over Ethernet config query response contains a "buffer count"
field reflecting the AoE target's capacity to buffer incoming AoE
commands.By taking the current value of this field into accound, we increase
performance throughput or avoid network congestion, when the value
has increased or decreased, respectively.Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Dropped transmits are not common, but when they do occur, increasing
the transmit queue length often helps.Signed-off-by: Ed Cashin
Cc: Dan Carpenter
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Signed-off-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The context feature of sparse is used with the Linux kernel sources to
check for imbalanced uses of locks. Document the annotations defined in
include/linux/compiler.h that tell sparse what to expect when a lock is
held on function entry, exit, or both.Signed-off-by: Ed Cashin
Reviewed-by: Josh Triplett
Acked-by: Christopher Li
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
linux/compiler.h has macros to denote functions that acquire or release
locks, but not to denote functions called with a lock held that return
with the lock still held. Add a __must_hold macro to cover that case.Signed-off-by: Josh Triplett
Reported-by: Ed Cashin
Tested-by: Ed Cashin
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Since commit 1cdcbec1a337 ("CRED: Neuter sys_capset()")
is_container_init() has no callers.Signed-off-by: Gao feng
Cc: David Howells
Acked-by: Serge Hallyn
Cc: James Morris
Cc: "Eric W. Biederman"
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
To avoid an explosion of request_module calls on a chain of abusive
scripts, fail maximum recursion with -ELOOP instead of -ENOEXEC. As soon
as maximum recursion depth is hit, the error will fail all the way back
up the chain, aborting immediately.This also has the side-effect of stopping the user's shell from attempting
to reexecute the top-level file as a shell script. As seen in the
dash source:if (cmd != path_bshell && errno == ENOEXEC) {
*argv-- = cmd;
*argv = cmd = path_bshell;
goto repeat;
}The above logic was designed for running scripts automatically that lacked
the "#!" header, not to re-try failed recursion. On a legitimate -ENOEXEC,
things continue to behave as the shell expects.Additionally, when tracking recursion, the binfmt handlers should not be
involved. The recursion being tracked is the depth of calls through
search_binary_handler(), so that function should be exclusively responsible
for tracking the depth.Signed-off-by: Kees Cook
Cc: halfdog
Cc: P J P
Cc: Alexander Viro
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
We display a list of supplementary group for each process in
/proc//status. However, we show only the first 32 groups, not all of
them.Although this is rare, but sometimes processes do have more than 32
supplementary groups, and this kernel limitation breaks user-space apps
that rely on the group list in /proc//status.Number 32 comes from the internal NGROUPS_SMALL macro which defines the
length for the internal kernel "small" groups buffer. There is no
apparent reason to limit to this value.This patch removes the 32 groups printing limit.
The Linux kernel limits the amount of supplementary groups by NGROUPS_MAX,
which is currently set to 65536. And this is the maximum count of groups
we may possibly print.Signed-off-by: Artem Bityutskiy
Acked-by: Serge E. Hallyn
Acked-by: Kees Cook
Cc:
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
It is currently impossible to examine the state of seccomp for a given
process. While attaching with gdb and attempting "call
prctl(PR_GET_SECCOMP,...)" will work with some situations, it is not
reliable. If the process is in seccomp mode 1, this query will kill the
process (prctl not allowed), if the process is in mode 2 with prctl not
allowed, it will similarly be killed, and in weird cases, if prctl is
filtered to return errno 0, it can look like seccomp is disabled.When reviewing the state of running processes, there should be a way to
externally examine the seccomp mode. ("Did this build of Chrome end up
using seccomp?" "Did my distro ship ssh with seccomp enabled?")This adds the "Seccomp" line to /proc/$pid/status.
Signed-off-by: Kees Cook
Reviewed-by: Cyrill Gorcunov
Cc: Andrea Arcangeli
Cc: James Morris
Acked-by: Serge E. Hallyn
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
During c/r sessions we've found that there is no way at the moment to
fetch some VMA associated flags, such as mlock() and madvise().This leads us to a problem -- we don't know if we should call for mlock()
and/or madvise() after restore on the vma area we're bringing back to
life.This patch intorduces a new field into "smaps" output called VmFlags,
where all set flags associated with the particular VMA is shown as two
letter mnemonics.[ Strictly speaking for c/r we only need mlock/madvise bits but it has been
said that providing just a few flags looks somehow inconsistent. So all
flags are here now. ]This feature is made available on CONFIG_CHECKPOINT_RESTORE=n kernels, as
other applications may start to use these fields.The data is encoded in a somewhat awkward two letters mnemonic form, to
encourage userspace to be prepared for fields being added or removed in
the future.[a.p.zijlstra@chello.nl: props to use for_each_set_bit]
[sfr@canb.auug.org.au: props to use array instead of struct]
[akpm@linux-foundation.org: overall redesign and simplification]
[akpm@linux-foundation.org: remove unneeded braces per sfr, avoid using bloaty for_each_set_bit()]
Signed-off-by: Cyrill Gorcunov
Cc: Pavel Emelyanov
Cc: Peter Zijlstra
Cc: Stephen Rothwell
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Without this patch it is really hard to interpret a bounding set, if
CAP_LAST_CAP is unknown for a current kernel.Non-existant capabilities can not be deleted from a bounding set with help
of prctl.E.g.: Here are two examples without/with this patch.
CapBnd: ffffffe0fdecffff
CapBnd: 00000000fdecffffI suggest to hide non-existent capabilities. Here is two reasons.
* It's logically and easier for using.
* It helps to checkpoint-restore capabilities of tasks, because tasks
can be restored on another kernel, where CAP_LAST_CAP is bigger.Signed-off-by: Andrew Vagin
Cc: Andrew G. Morgan
Reviewed-by: Serge E. Hallyn
Cc: Pavel Emelyanov
Reviewed-by: Kees Cook
Cc: KAMEZAWA Hiroyuki
Cc: James Morris
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Ptrace jailers want to be sure that the tracee can never escape
from the control. However if the tracer dies unexpectedly the
tracee continues to run in potentially unsafe mode.Add the new ptrace option PTRACE_O_EXITKILL. If the tracer exits
it sends SIGKILL to every tracee which has this bit set.Note that the new option is not equal to the last-option << 1. Because
currently all options have an event, and the new one starts the eventless
group. It uses the random 20 bit, so we have the room for 12 more events,
but we can also add the new eventless options below this one.Suggested by Amnon Shiloh.
Signed-off-by: Oleg Nesterov
Tested-by: Amnon Shiloh
Cc: Denys Vlasenko
Cc: Michael Kerrisk
Cc: Serge Hallyn
Cc: Chris Evans
Cc: David Howells
Cc: "Eric W. Biederman"
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Update the documentation for simple_strto* to reflect that it has been
obsoleted and advise the usage of kstrto*.Signed-off-by: Eldad Zack
Cc: J. Bruce Fields
Cc: Joe Perches
Cc: Randy Dunlap
Cc: Alexey Dobriyan
Cc: Rob Landley
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
As Bruce Fields pointed out, kstrto* is currently lacking kerneldoc
comments. This patch adds kerneldoc comments to common variants of
kstrto*: kstrto(u)l, kstrto(u)ll and kstrto(u)int.Signed-off-by: Eldad Zack
Cc: J. Bruce Fields
Cc: Joe Perches
Cc: Randy Dunlap
Cc: Alexey Dobriyan
Cc: Rob Landley
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
keys-ecryptfs.txt was missing from 00-INDEX.
Signed-off-by: Jarkko Sakkinen
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds
