Michael Leun reported that running parallel opens on a fuse filesystem
can trigger a "kernel BUG at mm/truncate.c:475"
Gurudas Pai reported the same bug on NFS.
The reason is, unmap_mapping_range() is not prepared for more than
one concurrent invocation per inode. For example:
thread1: going through a big range, stops in the middle of a vma and
stores the restart address in vm_truncate_count.
thread2: comes in with a small (e.g. single page) unmap request on
the same vma, somewhere before restart_address, finds that the
vma was already unmapped up to the restart address and happily
returns without doing anything.
Another scenario would be two big unmap requests, both having to
restart the unmapping and each one setting vm_truncate_count to its
own value. This could go on forever without any of them being able to
finish.
Truncate and hole punching already serialize with i_mutex. Other
callers of unmap_mapping_range() do not, and it's difficult to get
i_mutex protection for all callers. In particular ->d_revalidate(),
which calls invalidate_inode_pages2_range() in fuse, may be called
with or without i_mutex.
This patch adds a new mutex to 'struct address_space' to prevent
running multiple concurrent unmap_mapping_range() on the same mapping.
[ We'll hopefully get rid of all this with the upcoming mm
preemptibility series by Peter Zijlstra, the "mm: Remove i_mmap_mutex
lockbreak" patch in particular. But that is for 2.6.39 ]
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz> Reported-by: Michael Leun <lkml20101129@newton.leun.net> Reported-by: Gurudas Pai <gurudas.pai@oracle.com> Tested-by: Gurudas Pai <gurudas.pai@oracle.com> Acked-by: Hugh Dickins <hughd@google.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Consider this scenario: When the size of the first received udp packet
is bigger than the receive buffer, MSG_TRUNC bit is set in msg->msg_flags.
However, if checksum error happens and this is a blocking socket, it will
goto try_again loop to receive the next packet. But if the size of the
next udp packet is smaller than receive buffer, MSG_TRUNC flag should not
be set, but because MSG_TRUNC bit is not cleared in msg->msg_flags before
receive the next packet, MSG_TRUNC is still set, which is wrong.
Fix this problem by clearing MSG_TRUNC flag when starting over for a
new packet.
Signed-off-by: Xufeng Zhang <xufeng.zhang@windriver.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
udpv6_recvmsg() function is not using the correct variable to determine
whether or not the socket is in non-blocking operation, this will lead
to unexpected behavior when a UDP checksum error occurs.
Consider a non-blocking udp receive scenario: when udpv6_recvmsg() is
called by sock_common_recvmsg(), MSG_DONTWAIT bit of flags variable in
udpv6_recvmsg() is cleared by "flags & ~MSG_DONTWAIT" in this call:
But it will always go to try_again as MSG_DONTWAIT has been cleared
from flags at call time -- only noblock contains the original value
of MSG_DONTWAIT, so the test should be:
if (noblock)
return -EAGAIN;
This is also consistent with what the ipv4/udp code does.
Signed-off-by: Xufeng Zhang <xufeng.zhang@windriver.com> Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Check against mistakenly passing in IPv6 addresses (which would result
in an INADDR_ANY bind) or similar incompatible sockaddrs.
Signed-off-by: Marcus Meissner <meissner@suse.de> Cc: Reinhard Max <max@suse.de> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
In 2.6.27, commit 393e52e33c6c2 (packet: deliver VLAN TCI to userspace)
added a small information leak.
Add padding field and make sure its zeroed before copy to user.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
A mis-configured filter can spam the logs with lots of stack traces.
Rate-limit the warnings and add printout of the bogus filter information.
Original-patch-by: Ben Greear <greearb@candelatech.com> Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There is a bug in free_unnecessary_pages() that causes it to
attempt to free too many pages in some cases, which triggers the
BUG_ON() in memory_bm_clear_bit() for copy_bm. Namely, if
count_data_pages() is initially greater than alloc_normal, we get
to_free_normal equal to 0 and "save" greater from 0. In that case,
if the sum of "save" and count_highmem_pages() is greater than
alloc_highmem, we subtract a positive number from to_free_normal.
Hence, since to_free_normal was 0 before the subtraction and is
an unsigned int, the result is converted to a huge positive number
that is used as the number of pages to free.
Fix this bug by checking if to_free_normal is actually greater
than or equal to the number we're going to subtract from it.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Reported-and-tested-by: Matthew Garrett <mjg@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There is a problem in hibernate_preallocate_memory() that it calls
preallocate_image_memory() with an argument that may be greater than
the total number of available non-highmem memory pages. If that's
the case, the OOM condition is guaranteed to trigger, which in turn
can cause significant slowdown to occur during hibernation.
To avoid that, make preallocate_image_memory() adjust its argument
before calling preallocate_image_pages(), so that the total number of
saveable non-highem pages left is not less than the minimum size of
a hibernation image. Change hibernate_preallocate_memory() to try to
allocate from highmem if the number of pages allocated by
preallocate_image_memory() is too low.
Modify free_unnecessary_pages() to take all possible memory
allocation patterns into account.
Reported-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Tested-by: M. Vefa Bicakci <bicave@superonline.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
A malicious user or buggy application can inject code and trigger an
infinite loop in inet_diag_bc_audit()
Also make sure each instruction is aligned on 4 bytes boundary, to avoid
unaligned accesses.
Reported-by: Dan Rosenberg <drosenberg@vsecurity.com> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This will let us use it on a nlmsghdr stored inside a netlink_callback.
Signed-off-by: Nelson Elhage <nelhage@ksplice.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
With glibc 2.11 or later that was built with --enable-multi-arch, the UML
link fails with undefined references to __rel_iplt_start and similar
symbols. In recent binutils, the default linker script defines these
symbols (see ld --verbose). Fix the UML linker scripts to match the new
defaults for these sections.
Signed-off-by: Roland McGrath <roland@redhat.com> Cc: Jeff Dike <jdike@addtoit.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as1465) continues implementation of the policy that errors
during suspend or hibernation should not prevent the system from going
to sleep.
In this case, failure to turn on the Suspend feature for a hub port
shouldn't be reported as an error. There are situations where this
does actually occur (such as when the device plugged into that port
was disconnected in the recent past), and it turns out to be harmless.
There's no reason for it to prevent a system sleep.
Also, don't allow the hub driver to fail a system suspend if the
downstream ports aren't all suspended. This is also harmless (and
should never happen, given the change mentioned above); printing a
warning message in the kernel log is all we really need to do.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as1464) implements the recommended policy that most errors
during suspend or hibernation should not prevent the system from going
to sleep. In particular, failure to suspend a USB driver or a USB
device should not prevent the sleep from succeeding:
Failure to suspend a device won't matter, because the device will
automatically go into suspend mode when the USB bus stops carrying
packets. (This might be less true for USB-3.0 devices, but let's not
worry about them now.)
Failure of a driver to suspend might lead to trouble later on when the
system wakes up, but it isn't sufficient reason to prevent the system
from going to sleep.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently a single process may register exit handlers unlimited times.
It may lead to a bloated listeners chain and very slow process
terminations.
Eg after 10KK sent TASKSTATS_CMD_ATTR_REGISTER_CPUMASKs ~300 Mb of
kernel memory is stolen for the handlers chain and "time id" shows 2-7
seconds instead of normal 0.003. It makes it possible to exhaust all
kernel memory and to eat much of CPU time by triggerring numerous exits
on a single CPU.
The patch limits the number of times a single process may register
itself on a single CPU to one.
One little issue is kept unfixed - as taskstats_exit() is called before
exit_files() in do_exit(), the orphaned listener entry (if it was not
explicitly deregistered) is kept until the next someone's exit() and
implicit deregistration in send_cpu_listeners(). So, if a process
registered itself as a listener exits and the next spawned process gets
the same pid, it would inherit taskstats attributes.
The message hints that disc_data_lock is aquired with softirqs disabled,
but does not itself disable softirqs, which can in rare circumstances
lead to a deadlock.
The same problem is present in the 6pack driver, this patch fixes both
by using write_lock_bh instead of write_lock.
Reported-by: Bernard F6BVP <f6bvp@free.fr> Tested-by: Bernard F6BVP <f6bvp@free.fr> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Ralf Baechle<ralf@linux-mips.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
If a device fails in a way that causes pending request to take a while
to complete, md will not be able to immediately remove it from the
array in remove_and_add_spares.
It will then incorrectly look like a spare device and md will try to
recover it even though it is failed.
This leads to a recovery process starting and instantly aborting over
and over again.
We should check if the device is faulty before considering it to be a
spare. This will avoid trying to start a recovery that cannot
proceed.
This bug was introduced in 2.6.26 so that patch is suitable for any
kernel since then.
Since we are modifying this RCU pointer, we need to hold
the lock protecting it around it.
This fixes a potential reuse and double free of a cfq
io_context structure. The bug has been in CFQ for a long
time, it hit very few people but those it did hit seemed
to see it a lot.
Order of initialization look like this:
...
debugobjects
kmemleak
...(lots of other subsystems)...
workqueues (through early initcall)
...
debugobjects use schedule_work for batch freeing of its data and kmemleak
heavily use debugobjects, so when it comes to freeing and workqueues were
not initialized yet, kernel crashes:
When 1GB hugepages are allocated on a system, free(1) reports less
available memory than what really is installed in the box. Also, if the
total size of hugepages allocated on a system is over half of the total
memory size, CommitLimit becomes a negative number.
The problem is that gigantic hugepages (order > MAX_ORDER) can only be
allocated at boot with bootmem, thus its frames are not accounted to
'totalram_pages'. However, they are accounted to hugetlb_total_pages()
What happens to turn CommitLimit into a negative number is this
calculation, in fs/proc/meminfo.c:
A similar calculation occurs in __vm_enough_memory() in mm/mmap.c.
Also, every vm statistic which depends on 'totalram_pages' will render
confusing values, as if system were 'missing' some part of its memory.
Impact of this bug:
When gigantic hugepages are allocated and sysctl_overcommit_memory ==
OVERCOMMIT_NEVER. In a such situation, __vm_enough_memory() goes through
the mentioned 'allowed' calculation and might end up mistakenly returning
-ENOMEM, thus forcing the system to start reclaiming pages earlier than it
would be ususal, and this could cause detrimental impact to overall
system's performance, depending on the workload.
Besides the aforementioned scenario, I can only think of this causing
annoyances with memory reports from /proc/meminfo and free(1).
[akpm@linux-foundation.org: standardize comment layout] Reported-by: Russ Anderson <rja@sgi.com> Signed-off-by: Rafael Aquini <aquini@linux.com> Acked-by: Russ Anderson <rja@sgi.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Christoph Lameter <cl@linux.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We would free the proper number of curves, but in the wrong
slots, due to a missing level of indirection through
the pdgain_idx table.
It's simpler just to try to free all four slots, so do that.
Signed-off-by: Bob Copeland <me@bobcopeland.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When opening /dev/snapshot device, snapshot_open() creates memory
bitmaps which are freed in snapshot_release(). But if any of the
callbacks called by pm_notifier_call_chain() returns NOTIFY_BAD, open()
fails, snapshot_release() is never called and bitmaps are not freed.
Next attempt to open /dev/snapshot then triggers BUG_ON() check in
create_basic_memory_bitmaps(). This happens e.g. when vmwatchdog module
is active on s390x.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
While trying to switch a UAS device from the BOT configuration to the UAS
configuration via the bConfigurationValue file, Tanya ran into an issue in
the USB core. usb_disable_device() sets entries in udev->ep_out and
udev->ep_out to NULL, but doesn't call into the xHCI bandwidth management
functions to remove the BOT configuration endpoints from the xHCI host's
internal structures.
The USB core would then attempt to add endpoints for the UAS
configuration, and some of the endpoints had the same address as endpoints
in the BOT configuration. The xHCI driver blindly added the endpoints
again, but the xHCI host controller rejected the Configure Endpoint
command because active endpoints were added without being dropped.
Make the xHCI driver reject calls to xhci_add_endpoint() that attempt to
add active endpoints without first calling xhci_drop_endpoint().
This should be backported to kernels as old as 2.6.31.
We restored tty_ldisc_wait_idle in 100eeae2c5c (TTY: restore
tty_ldisc_wait_idle). We used it in the ldisc changing path to fix the
case where there are tasks in n_tty_read waiting for data and somebody
tries to change ldisc.
Similar to the case above, there may be also tasks waiting in
n_tty_read while hangup is performed. As 65b770468e98 (tty-ldisc: turn
ldisc user count into a proper refcount) removed the wait-until-idle
from all paths, hangup path won't wait for them to disappear either
now. So add it back even to the hangup path.
There is a difference, we need uninterruptible sleep as there is
obviously HUP signal pending. So tty_ldisc_wait_idle now sleeps
without possibility to be interrupted. This is what original
tty_ldisc_wait_idle did. After the wait idle reintroduction
(100eeae2c5c), we have had interruptible sleeps for the ldisc changing
path. But as there is a 5s timeout anyway, we don't allow it to be
interrupted from now on. It's not worth the added complexity of
deciding what kind of sleep we want.
Before 65b770468e98 tty_ldisc_release was called also from
tty_ldisc_release. It is called from tty_release, so I don't think we
need to restore that one.
This is nicely reproducible after constifying the timing when
drivers/tty/n_tty.c is patched as follows ("TTY: ntty, add one more
sanity check" patch is needed to actually see it explode):
%% -1548,6 +1549,7 @@ static int n_tty_open(struct tty_struct *tty)
/* These are ugly. Currently a malloc failure here can panic */
if (!tty->read_buf) {
+ msleep(100);
tty->read_buf = kzalloc(N_TTY_BUF_SIZE, GFP_KERNEL);
if (!tty->read_buf)
return -ENOMEM;
%% -1785,6 +1788,7 @@ do_it_again:
break;
}
timeout = schedule_timeout(timeout);
+ msleep(20);
continue;
}
__set_current_state(TASK_RUNNING);
===== With a process: =====
while (1) {
int fd = open(argv[1], O_RDWR);
read(fd, buf, sizeof(buf));
close(fd);
}
===== and its child: =====
setsid();
while (1) {
int fd = open(tty, O_RDWR|O_NOCTTY);
ioctl(fd, TIOCSCTTY, 1);
vhangup();
close(fd);
usleep(100 * (10 + random() % 1000));
}
===== EOF =====
The clocksource watchdog code is interruptible and it has been
observed that this can trigger false positives which disable the TSC.
The reason is that an interrupt storm or a long running interrupt
handler between the read of the watchdog source and the read of the
TSC brings the two far enough apart that the delta is larger than the
unstable treshold. Move both reads into a short interrupt disabled
region to avoid that.
Reported-and-tested-by: Vernon Mauery <vernux@us.ibm.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We only need to set max_pfn_mapped to the last pfn mapped on x86_64 to
make sure that cleanup_highmap doesn't remove important mappings at
_end.
We don't need to do this on x86_32 because cleanup_highmap is not called
on x86_32. Besides lowering max_pfn_mapped on x86_32 has the unwanted
side effect of limiting the amount of memory available for the 1:1
kernel pagetable allocation.
This patch reverts the x86_32 part of the original patch.
Andrea Righi reported a case where an exiting task can race against
ksmd::scan_get_next_rmap_item (http://lkml.org/lkml/2011/6/1/742) easily
triggering a NULL pointer dereference in ksmd.
ksm_scan.mm_slot == &ksm_mm_head with only one registered mm
CPU 1 (__ksm_exit) CPU 2 (scan_get_next_rmap_item)
list_empty() is false
lock slot == &ksm_mm_head
list_del(slot->mm_list)
(list now empty)
unlock
lock
slot = list_entry(slot->mm_list.next)
(list is empty, so slot is still ksm_mm_head)
unlock
slot->mm == NULL ... Oops
Close this race by revalidating that the new slot is not simply the list
head again.
It turns out this is not what we want to have happen for the .32 and
.33-longterm kernels as it does not work properly at all.
This was reported by Gentoo, Arch, and Canonical developers as causing
problems for their users:
https://bugs.archlinux.org/task/24302
http://bugs.gentoo.org/show_bug.cgi?id=359445
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/796336
Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Gordon Malm <gengor@gentoo.org> Cc: Don Fry <donald.h.fry@intel.com> Cc: Wey-Yi Guy <wey-yi.w.guy@intel.com> Cc: Stanislaw Gruszka <sgruszka@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The mask indicates the bits one wants to zero out, so it needs to be
inverted before applying to the original TOS field.
Signed-off-by: Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The IPv6 header is not zeroed out in alloc_skb so we must initialize
it properly unless we want to see IPv6 packets with random TOS fields
floating around. The current implementation resets the flow label
but this could be changed if deemed necessary.
We stumbled upon this issue when trying to apply a mangle rule to
the RST packet generated by the REJECT target module.
Signed-off-by: Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When marking an inode reclaimable, a per-AG counter is increased, the
inode is tagged reclaimable in its per-AG tree, and, when this is the
first reclaimable inode in the AG, the AG entry in the per-mount tree
is also tagged.
When an inode is finally reclaimed, however, it is only deleted from
the per-AG tree. Neither the counter is decreased, nor is the parent
tree's AG entry untagged properly.
Since the tags in the per-mount tree are not cleared, the inode
shrinker iterates over all AGs that have had reclaimable inodes at one
point in time.
The counters on the other hand signal an increasing amount of slab
objects to reclaim. Since "70e60ce xfs: convert inode shrinker to
per-filesystem context" this is not a real issue anymore because the
shrinker bails out after one iteration.
But the problem was observable on a machine running v2.6.34, where the
reclaimable work increased and each process going into direct reclaim
eventually got stuck on the xfs inode shrinking path, trying to scan
several million objects.
Fix this by properly unwinding the reclaimable-state tracking of an
inode when it is reclaimed.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Alex Elder <aelder@sgi.com> Backported-by: Stefan Priebe <s.priebe@profihost.ag> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The old IDE cmd64x checks the status of the CNTRL register to see if
the ports are enabled before probing them. pata_cmd64x doesn't do
this, which causes a HPMC on parisc when it tries to poke at the
secondary port because apparently the BAR isn't wired up (and a
non-responding piece of memory causes a HPMC).
Fix this by porting the CNTRL register port detection logic from IDE
cmd64x. In addition, following converns from Alan Cox, add a check to
see if a mobility electronics bridge is the immediate parent and forgo
the check if it is (prevents problems on hotplug controllers).
Signed-off-by: James Bottomley <James.Bottomley@suse.de> Signed-off-by: Jeff Garzik <jgarzik@pobox.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The @bio->bi_phys_segments consists of active stripes count in the
lower 16 bits and processed stripes count in the upper 16 bits. So
logical-OR operator should be bitwise one.
This bug has been present since 2.6.27 and the fix is suitable for any
-stable kernel since then. Fortunately the bad code is only used on
error paths and is relatively unlikely to be hit.
Check pers->hot_remove_disk instead of pers->hot_add_disk in slot_store()
during disk removal. The linear personality only has ->hot_add_disk and
no ->hot_remove_disk, so that removing disk in the array resulted to
following kernel bug:
cpufreq_stats leaves behind its sysfs entries, which causes a panic
when something stumbled across them.
(Discovered by unloading cpufreq_stats while powertop was loaded).
Signed-off-by: Dave Jones <davej@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The lockdep warning below detects a possible A->B/B->A locking
dependency of mm->mmap_sem and dcookie_mutex. The order in
sync_buffer() is mm->mmap_sem/dcookie_mutex, while in
sys_lookup_dcookie() it is vice versa.
Fixing it in sys_lookup_dcookie() by unlocking dcookie_mutex before
copy_to_user().
oprofiled/4432 is trying to acquire lock:
(&mm->mmap_sem){++++++}, at: [<ffffffff810b444b>] might_fault+0x53/0xa3
but task is already holding lock:
(dcookie_mutex){+.+.+.}, at: [<ffffffff81124d28>] sys_lookup_dcookie+0x45/0x149
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
BugLink: https://launchpad.net/bugs/792712
The original reporter states that sound from the internal speakers is
inaudible until using the model=auto quirk. This symptom is due to an
existing quirk mask for 0x102802b* that uses the model=dell quirk. To
limit the possible regressions, leave the existing quirk mask but add
a higher priority specific mask for the reporter's PCI SSID.
Reported-and-tested-by: rodni hipp Signed-off-by: Daniel T Chen <crimsun@ubuntu.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Full-speed isoc endpoints specify interval in exponent based form in
frames, not microframes, so we need to adjust accordingly.
NEC xHCI host controllers will return an error code of 0x11 if a full
speed isochronous endpoint is added with the Interval field set to
something less than 3 (2^3 = 8 microframes, or one frame). It is
impossible for a full speed device to have an interval smaller than one
frame.
This was always an issue in the xHCI driver, but commit dfa49c4ad120a784ef1ff0717168aa79f55a483a "USB: xhci - fix math in
xhci_get_endpoint_interval()" removed the clamping of the minimum value
in the Interval field, which revealed this bug.
This needs to be backported to stable kernels back to 2.6.31.
Reported-by: Matt Evans <matt@ozlabs.org> Signed-off-by: Dmitry Torokhov <dtor@vmware.com> Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Protocol stall should not be fatal while reading port or hub status as it is
transient state. Currently hub EP0 STALL during port status read results in
failed device enumeration. This has been observed with ST-Ericsson (formerly
Philips) USB 2.0 Hub (04cc:1521) after connecting keyboard.
Some PCIe cards ship with a PCI-PCIe bridge which is not
visible as a PCI device in Linux. But the device-id of the
bridge is present in the IOMMU tables which causes a boot
crash in the IOMMU driver.
This patch fixes by removing these cards from the IOMMU
handling. This is a pure -stable fix, a real fix to handle
this situation appriatly will follow for the next merge
window.
The driver contains several loops counting on an u16 value
where the exit-condition is checked against variables that
can have values up to 0xffff. In this case the loops will
never exit. This patch fixed 3 such loops.
Unfortunatly there are systems where the AMD IOMMU does not
cover all devices. This breaks with the current driver as it
initializes the global dma_ops variable. This patch limits
the AMD IOMMU to the devices listed in the IVRS table fixing
DMA for devices not covered by the IOMMU.
On my x86_64 system with >4GB of ram and swiotlb instead of
a hardware iommu (because I have a VIA chipset), the call
to pci_set_dma_mask (see below) with 40bits returns an error.
But it seems that the radeon driver is designed to have
need_dma32 = true exactly if pci_set_dma_mask is called
with 32 bits and false if it is called with 40 bits.
I have read somewhere that the default are 32 bits. So if the
call fails I suppose that need_dma32 should be set to true.
And indeed the patch fixes the problem I have had before
and which I had described here:
http://choon.net/forum/read.php?21,106131,115940
Acked-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I found this while figuring out why gnome-shell would not run on my
Asus EeeBox PC EB1007. As a standalone "pc" this device cleary does not have
an internal panel, yet it claims it does. Add a quirk to fix this.
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Reviewed-by: Keith Packard <keithp@keithp.com> Signed-off-by: Keith Packard <keithp@keithp.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The main lock_is_held() user is lockdep_assert_held(), avoid false
assertions in lockdep_off() sections by unconditionally reporting the
lock is taken.
[ the reason this is important is a lockdep_assert_held() in ttwu()
which triggers a warning under lockdep_off() as in printk() which
can trigger another wakeup and lock up due to spinlock
recursion, as reported and heroically debugged by Arne Jansen ]
In both trigger_scan and sched_scan operations, we were checking for
the SSID length before assigning the value correctly. Since the
memory was just kzalloc'ed, the check was always failing and SSID with
over 32 characters were allowed to go through.
This was causing a buffer overflow when copying the actual SSID to the
proper place.
This bug has been there since 2.6.29-rc4.
Signed-off-by: Luciano Coelho <coelho@ti.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The following patch sets the MaxPayload setting to match the parent
reading when inserting a PCIE card into a hotplug slot. On our system,
the upstream bridge is set to 256, but when inserting a card, the card
setting defaults to 128. As soon as I/O is performed to the card it
starts receiving errors since the payload size is too small.
Al Viro observes that in the hugetlb case, handle_mm_fault() may return
a value of the kind ENOSPC when its caller is expecting a value of the
kind VM_FAULT_SIGBUS: fix alloc_huge_page()'s failure returns.
Whenever there is a channel width change from 40 Mhz to 20 Mhz,
the hardware is reconfigured to ht20. Meantime before doing
the rate control updation, the packets are being transmitted are
selected rate with IEEE80211_TX_RC_40_MHZ_WIDTH.
While transmitting ht40 rate packets in ht20 mode is causing
baseband panic with AR9003 based chips.
In certain circumstances, we can get an oops from a torn down device.
Most notably this is from CD roms trying to call scsi_ioctl. The root
cause of the problem is the fact that after scsi_remove_device() has
been called, the queue is fully torn down. This is actually wrong
since the queue can be used until the sdev release function is called.
Therefore, we add an extra reference to the queue which is released in
sdev->release, so the queue always exists.
The 'max_part' parameter controls the number of maximum partition
a nbd device can have. However if a user specifies very large
value it would exceed the limitation of device minor number and
can cause a kernel oops (or, at least, produce invalid device
nodes in some cases).
In addition, specifying large 'nbds_max' value causes same
problem for the same reason.
On my desktop, following command results to the kernel bug:
UBIFS leaks memory on error path in 'ubifs_jnl_update()' in case of write
failure because it forgets to free the 'struct ubifs_dent_node *dent' object.
Although the object is small, the alignment can make it large - e.g., 2KiB
if the min. I/O unit is 2KiB.
Sometimes VM asks the shrinker to return amount of objects it can shrink,
and we return the ubifs_clean_zn_cnt in that case. However, it is possible
that this counter is negative for a short period of time, due to the way
UBIFS TNC code updates it. And I can observe the following warnings sometimes:
shrink_slab: ubifs_shrinker+0x0/0x2b7 [ubifs] negative objects to delete nr=-8541616642706119788
This patch makes sure UBIFS never returns negative count of objects.
This patch (as1335) fixes a bug in scsi_sysfs_add_sdev(). Its callers
always remove the device if anything goes wrong, so it should never
remove the device.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: James Bottomley <James.Bottomley@suse.de> Cc: Hannes Reinecke <hare@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When an ICMPV6_PKT_TOOBIG message is received with a MTU below 1280,
all further packets include a fragment header.
Unlike regular defragmentation, conntrack also needs to "reassemble"
those fragments in order to obtain a packet without the fragment
header for connection tracking. Currently nf_conntrack_reasm checks
whether a fragment has either IP6_MF set or an offset != 0, which
makes it ignore those fragments.
Remove the invalid check and make reassembly handle fragment queues
containing only a single fragment.
There's a race window in xen_drop_mm_ref, where remote cpu may exit
dirty bitmap between the check on this cpu and the point where remote
cpu handles drop request. So in drop_other_mm_ref we need check
whether TLB state is still lazy before calling into leave_mm. This
bug is rarely observed in earlier kernel, but exaggerated by the
commit 831d52bc153971b70e64eccfbed2b232394f22f8
("x86, mm: avoid possible bogus tlb entries by clearing prev mm_cpumask after switching mm")
which clears bitmap after changing the TLB state. the call trace is as below:
Tested-by: Maoxiaoyun<tinnycloud@hotmail.com> Signed-off-by: Kevin Tian <kevin.tian@intel.com>
[v1: Fleshed out the git description a bit] Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
TI816X (common name for DM816x/C6A816x/AM389x family) devices configured
to boot as PCIe Endpoint have class code = 0. This makes kernel PCI bus
code to skip allocating BARs to these devices resulting into following
type of error when trying to enable them:
"Device 0000:01:00.0 not available because of resource collisions"
The device cannot be operated because of the above issue.
This patch adds a ID specific (TI VENDOR ID and 816X DEVICE ID based)
'early' fixup quirk to replace class code with
PCI_CLASS_MULTIMEDIA_VIDEO as class.
The TCP connection state code depends on the state_change() callback
being called when the SYN_SENT state is set. However the networking layer
doesn't actually call us back in that case.
When finding or allocating a ram disk device, brd_probe() did not take
partition numbers into account so that it can result to a different
device. Consider following example (I set CONFIG_BLK_DEV_RAM_COUNT=4
for simplicity) :
$ sudo modprobe brd max_part=15
$ ls -l /dev/ram*
brw-rw---- 1 root disk 1, 0 2011-05-25 15:41 /dev/ram0
brw-rw---- 1 root disk 1, 16 2011-05-25 15:41 /dev/ram1
brw-rw---- 1 root disk 1, 32 2011-05-25 15:41 /dev/ram2
brw-rw---- 1 root disk 1, 48 2011-05-25 15:41 /dev/ram3
$ sudo mknod /dev/ram4 b 1 64
$ sudo dd if=/dev/zero of=/dev/ram4 bs=4k count=256
256+0 records in
256+0 records out 1048576 bytes (1.0 MB) copied, 0.00215578 s, 486 MB/s
namhyung@leonhard:linux$ ls -l /dev/ram*
brw-rw---- 1 root disk 1, 0 2011-05-25 15:41 /dev/ram0
brw-rw---- 1 root disk 1, 16 2011-05-25 15:41 /dev/ram1
brw-rw---- 1 root disk 1, 32 2011-05-25 15:41 /dev/ram2
brw-rw---- 1 root disk 1, 48 2011-05-25 15:41 /dev/ram3
brw-r--r-- 1 root root 1, 64 2011-05-25 15:45 /dev/ram4
brw-rw---- 1 root disk 1, 1024 2011-05-25 15:44 /dev/ram64
After this patch, /dev/ram4 - instead of /dev/ram64 - was
accessed correctly.
In addition, 'range' passed to blk_register_region() should
include all range of dev_t that RAMDISK_MAJOR can address.
It does not need to be limited by partition numbers unless
'rd_nr' param was specified.
The 'max_part' parameter controls the number of maximum partition
a brd device can have. However if a user specifies very large
value it would exceed the limitation of device minor number and
can cause a kernel panic (or, at least, produce invalid device
nodes in some cases).
On my desktop system, following command kills the kernel. On qemu,
it triggers similar oops but the kernel was alive:
$ sudo modprobe brd max_part=100000
BUG: unable to handle kernel NULL pointer dereference at 0000000000000058
IP: [<ffffffff81110a9a>] sysfs_create_dir+0x2d/0xae
PGD 7af1067 PUD 7b19067 PMD 0
Oops: 0000 [#1] SMP
last sysfs file:
CPU 0
Modules linked in: brd(+)
It's currently exposed only through /proc which, besides requiring
screen-scraping, doesn't allow userspace to distinguish between two
identical ATM adapters with different ATM indexes. The ATM device index
is required when using PPPoATM on a system with multiple ATM adapters.
Signed-off-by: Dan Williams <dcbw@redhat.com> Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com> Tested-by: David Woodhouse <dwmw2@infradead.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The v6 and v7 implementations of flush_kern_dcache_area do not align
the passed MVA to the size of a cacheline in the data cache. If a
misaligned address is used, only a subset of the requested area will
be flushed. This has been observed to cause failures in SMP boot where
the secondary_data initialised by the primary CPU is not cacheline
aligned, causing the secondary CPUs to read incorrect values for their
pgd and stack pointers.
This patch ensures that the base address is cacheline aligned before
flushing the d-cache.
Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch adds a check that a block device has a request function
defined before it is used. Otherwise, misconfiguration can cause an oops.
Because we are allowing devices with zero size e.g. an offline multipath
device as in commit 2cd54d9bedb79a97f014e86c0da393416b264eb3
("dm: allow offline devices") there needs to be an additional check
to ensure devices are initialised. Some block devices, like a loop
device without a backing file, exist but have no request function.
Reproducer is trivial: dm-mirror on unbound loop device
(no backing file on loop devices)
Cpuidle menu governor is using u32 as a temporary datatype for storing
nanosecond values which wrap around at 4.294 seconds. This causes errors
in predicted sleep times resulting in higher than should be C state
selection and increased power consumption. This also breaks cpuidle
state residency statistics.
Signed-off-by: Tero Kristo <tero.kristo@nokia.com> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
i8k uses lahf to read the flag register in 64-bit code; early x86-64
CPUs, however, lack this instruction and we get an invalid opcode
exception at runtime.
Use pushf to load the flag register into the stack instead.
Signed-off-by: Luca Tettamanti <kronos.it@gmail.com> Reported-by: Jeff Rickman <jrickman@myamigos.us> Tested-by: Jeff Rickman <jrickman@myamigos.us> Tested-by: Harry G McGavran Jr <w5pny@arrl.net> Cc: Massimo Dal Zotto <dz@debian.org> Signed-off-by: Jean Delvare <khali@linux-fr.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When re-mounting from R/O mode to R/W mode and the LEB count in the superblock
is not up-to date, because for the underlying UBI volume became larger, we
re-write the superblock. We allocate RAM for these purposes, but never free it.
So this is a memory leak, although very rare one.
The buffers allocated while encrypting and decrypting long filenames can
sometimes straddle two pages. In this situation, virt_to_scatterlist()
will return -ENOMEM, causing the operation to fail and the user will get
scary error messages in their logs:
kernel: ecryptfs_write_tag_70_packet: Internal error whilst attempting
to convert filename memory to scatterlist; expected rc = 1; got rc =
[-12]. block_aligned_filename_size = [272]
kernel: ecryptfs_encrypt_filename: Error attempting to generate tag 70
packet; rc = [-12]
kernel: ecryptfs_encrypt_and_encode_filename: Error attempting to
encrypt filename; rc = [-12]
kernel: ecryptfs_lookup: Error attempting to encrypt and encode
filename; rc = [-12]
The solution is to allow up to 2 scatterlist entries to be used.
Reported-by: Mark Davis <marked86@gmail.com> Signed-off-by: Christian Lamparter <chunkeey@googlemail.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
introduced a bug. The USB 2.0 spec says that full speed isochronous endpoints'
bInterval must be decoded as an exponent to a power of two (e.g. interval =
2^(bInterval - 1)). Full speed interrupt endpoints, on the other hand, don't
use exponents, and the interval in frames is encoded straight into bInterval.
Dmitry's patch was supposed to fix up the full speed isochronous to parse
bInterval as an exponent, but instead it changed the *interrupt* endpoint
bInterval decoding. The isochronous endpoint encoding was the same.
This caused full speed devices with interrupt endpoints (including mice, hubs,
and USB to ethernet devices) to fail under NEC 0.96 xHCI host controllers:
[ 100.909818] xhci_hcd 0000:06:00.0: add ep 0x83, slot id 1, new drop flags = 0x0, new add flags = 0x99, new slot info = 0x38100000
[ 100.909821] xhci_hcd 0000:06:00.0: xhci_check_bandwidth called for udev ffff88011f0ea000
...
[ 100.910187] xhci_hcd 0000:06:00.0: ERROR: unexpected command completion code 0x11.
[ 100.910190] xhci_hcd 0000:06:00.0: xhci_reset_bandwidth called for udev ffff88011f0ea000
When the interrupt endpoint was added and a Configure Endpoint command was
issued to the host, the host controller would return a very odd error message
(0x11 means "Slot Not Enabled", which isn't true because the slot was enabled).
Probably the host controller was getting very confused with the bad encoding.
Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Cc: Dmitry Torokhov <dtor@vmware.com> Reported-by: Thomas Lindroth <thomas.lindroth@gmail.com> Tested-by: Thomas Lindroth <thomas.lindroth@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
composite.c always sets req->length to zero
and expects function driver's setup handlers
to return the amount of bytes to be used
on req->length. If we test against req->length
w_length will always be greater than req->length
thus making us always stall that particular
SEND_ENCAPSULATED_COMMAND request.
Tested against a Windows XP SP3.
Signed-off-by: Felipe Balbi <balbi@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch fixes a problem where data received from the gps is sometimes
transferred incompletely to the serial port. If used in native mode now
all data received via the bulk queue will be forwarded to the serial
port.
Signed-off-by: Hermann Kneissel <herkne@gmx.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When finding or allocating a loop device, loop_probe() did not take
partition numbers into account so that it can result to a different
device. Consider following example:
$ sudo modprobe loop max_part=15
$ ls -l /dev/loop*
brw-rw---- 1 root disk 7, 0 2011-05-24 22:16 /dev/loop0
brw-rw---- 1 root disk 7, 16 2011-05-24 22:16 /dev/loop1
brw-rw---- 1 root disk 7, 32 2011-05-24 22:16 /dev/loop2
brw-rw---- 1 root disk 7, 48 2011-05-24 22:16 /dev/loop3
brw-rw---- 1 root disk 7, 64 2011-05-24 22:16 /dev/loop4
brw-rw---- 1 root disk 7, 80 2011-05-24 22:16 /dev/loop5
brw-rw---- 1 root disk 7, 96 2011-05-24 22:16 /dev/loop6
brw-rw---- 1 root disk 7, 112 2011-05-24 22:16 /dev/loop7
$ sudo mknod /dev/loop8 b 7 128
$ sudo losetup /dev/loop8 ~/temp/disk-with-3-parts.img
$ sudo losetup -a
/dev/loop128: [0805]:278201 (/home/namhyung/temp/disk-with-3-parts.img)
$ ls -l /dev/loop*
brw-rw---- 1 root disk 7, 0 2011-05-24 22:16 /dev/loop0
brw-rw---- 1 root disk 7, 16 2011-05-24 22:16 /dev/loop1
brw-rw---- 1 root disk 7, 2048 2011-05-24 22:18 /dev/loop128
brw-rw---- 1 root disk 7, 2049 2011-05-24 22:18 /dev/loop128p1
brw-rw---- 1 root disk 7, 2050 2011-05-24 22:18 /dev/loop128p2
brw-rw---- 1 root disk 7, 2051 2011-05-24 22:18 /dev/loop128p3
brw-rw---- 1 root disk 7, 32 2011-05-24 22:16 /dev/loop2
brw-rw---- 1 root disk 7, 48 2011-05-24 22:16 /dev/loop3
brw-rw---- 1 root disk 7, 64 2011-05-24 22:16 /dev/loop4
brw-rw---- 1 root disk 7, 80 2011-05-24 22:16 /dev/loop5
brw-rw---- 1 root disk 7, 96 2011-05-24 22:16 /dev/loop6
brw-rw---- 1 root disk 7, 112 2011-05-24 22:16 /dev/loop7
brw-r--r-- 1 root root 7, 128 2011-05-24 22:17 /dev/loop8
After this patch, /dev/loop8 - instead of /dev/loop128 - was
accessed correctly.
In addition, 'range' passed to blk_register_region() should
include all range of dev_t that LOOP_MAJOR can address. It does
not need to be limited by partition numbers unless 'max_loop'
param was specified.
The 'max_part' parameter controls the number of maximum partition
a loop block device can have. However if a user specifies very
large value it would exceed the limitation of device minor number
and can cause a kernel panic (or, at least, produce invalid
device nodes in some cases).
On my desktop system, following command kills the kernel. On qemu,
it triggers similar oops but the kernel was alive:
$ sudo modprobe loop max_part0000
------------[ cut here ]------------
kernel BUG at /media/Linux_Data/project/linux/fs/sysfs/group.c:65!
invalid opcode: 0000 [#1] SMP
last sysfs file:
CPU 0
Modules linked in: loop(+)
I believe I found a problem in __alloc_pages_slowpath, which allows a
process to get stuck endlessly looping, even when lots of memory is
available.
Running an I/O and memory intensive stress-test I see a 0-order page
allocation with __GFP_IO and __GFP_WAIT, running on a system with very
little free memory. Right about the same time that the stress-test gets
killed by the OOM-killer, the utility trying to allocate memory gets stuck
in __alloc_pages_slowpath even though most of the systems memory was freed
by the oom-kill of the stress-test.
The utility ends up looping from the rebalance label down through the
wait_iff_congested continiously. Because order=0,
__alloc_pages_direct_compact skips the call to get_page_from_freelist.
Because all of the reclaimable memory on the system has already been
reclaimed, __alloc_pages_direct_reclaim skips the call to
get_page_from_freelist. Since there is no __GFP_FS flag, the block with
__alloc_pages_may_oom is skipped. The loop hits the wait_iff_congested,
then jumps back to rebalance without ever trying to
get_page_from_freelist. This loop repeats infinitely.
The test case is pretty pathological. Running a mix of I/O stress-tests
that do a lot of fork() and consume all of the system memory, I can pretty
reliably hit this on 600 nodes, in about 12 hours. 32GB/node.
Signed-off-by: Andrew Barry <abarry@cray.com> Signed-off-by: Minchan Kim <minchan.kim@gmail.com> Reviewed-by: Rik van Riel<riel@redhat.com> Acked-by: Mel Gorman <mgorman@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There are no signs of a dmic at node 0x0b, so the user is left with
an additional internal mic which does not exist. This commit removes
that non-existing mic.
BugLink: http://bugs.launchpad.net/bugs/731706 Reported-by: James Page <james.page@canonical.com> Signed-off-by: David Henningsson <david.henningsson@canonical.com> Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When the clocksource is not a multiple of HZ, the clock will be off. For
acpi_pm, HZ=1000 the error is 127.111 ppm:
The rounding of cycle_interval ends up generating a false error term in
ntp_error accumulation since xtime_interval is not exactly 1/HZ. So, we
subtract out the error caused by the rounding.
This has been visible since 2.6.32-rc2
commit a092ff0f90cae22b2ac8028ecd2c6f6c1a9e4601
time: Implement logarithmic time accumulation
That commit raised NTP_INTERVAL_FREQ and exposed the rounding error.
testing tool: http://n1.taur.dk/permanent/testpmt.c
Also tested with ntpd and a frequency counter.
Signed-off-by: Kasper Pedersen <kkp2010@kasperkp.dk> Acked-by: john stultz <johnstul@us.ibm.com> Cc: John Kacur <jkacur@redhat.com> Cc: Clark Williams <williams@redhat.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Move the smp_rmb after cpu_relax loop in read_seqlock and add
ACCESS_ONCE to make sure the test and return are consistent.
A multi-threaded core in the lab didn't like the update
from 2.6.35 to 2.6.36, to the point it would hang during
boot when multiple threads were active. Bisection showed af5ab277ded04bd9bc6b048c5a2f0e7d70ef0867 (clockevents:
Remove the per cpu tick skew) as the culprit and it is
supported with stack traces showing xtime_lock waits including
tick_do_update_jiffies64 and/or update_vsyscall.
Experimentation showed the combination of cpu_relax and smp_rmb
was significantly slowing the progress of other threads sharing
the core, and this patch is effective in avoiding the hang.
A theory is the rmb is affecting the whole core while the
cpu_relax is causing a resource rebalance flush, together they
cause an interfernce cadance that is unbroken when the seqlock
reader has interrupts disabled.
At first I was confused why the refactor in 3c22cd5709e8143444a6d08682a87f4c57902df3 (kernel: optimise
seqlock) didn't affect this patch application, but after some
study that affected seqcount not seqlock. The new seqcount was
not factored back into the seqlock. I defer that the future.
While the removal of the timer interrupt offset created
contention for the xtime lock while a cpu does the
additonal work to update the system clock, the seqlock
implementation with the tight rmb spin loop goes back much
further, and is just waiting for the right trigger.
Signed-off-by: Milton Miller <miltonm@bga.com> Cc: <linuxppc-dev@lists.ozlabs.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Andi Kleen <andi@firstfloor.org> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Anton Blanchard <anton@samba.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Acked-by: Eric Dumazet <eric.dumazet@gmail.com> Link: http://lkml.kernel.org/r/%3Cseqlock-rmb%40mdm.bga.com%3E Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
As Ben Hutchings discovered [1], the patch for CVE-2011-1017 (buffer
overflow in ldm_frag_add) is not sufficient. The original patch in
commit c340b1d64000 ("fs/partitions/ldm.c: fix oops caused by corrupted
partition table") does not consider that, for subsequent fragments,
previously allocated memory is used.
[1] http://lkml.org/lkml/2011/5/6/407
Reported-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Timo Warns <warns@pre-sense.de> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
projects / karo-tx-linux.git / log