New connection tracking entries are inserted into the hash before they
are fully set up, namely the CONFIRMED bit is not set and the timer not
started yet. This can theoretically lead to a race with timer, which
would set the timeout value to a relative value, most likely already in
the past.
Perform hash insertion as the final step to fix this.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Commit 31207dab7d2e63795eb15823947bd2f7025b08e2
"Fix incorrect allocation of interrupt rev-map"
introduced a regression crashing on boot on machines using
a "DCR" based MPIC, such as the Cell blades.
The reason is that the irq host data structure is initialized
much later as a result of that patch, causing our calls to
mpic_map() do be done before we have a host setup.
Unfortunately, this breaks _mpic_map_dcr() which uses the
mpic->irqhost to get to the device node.
This fixes it by, instead, passing the device node explicitely
to mpic_map().
In testing a backport of the write_begin/write_end AOPs, a 10% re-read
regression was noticed when running iozone. This regression was
introduced because the old AOPs would always do a mark_page_accessed(page)
after the commit_write, but when the new AOPs where introduced, the only
place this was kept was in pagecache_write_end().
This patch does the same thing in the generic case as what is done in
pagecache_write_end(), which is just to mark the page accessed before we
do write_end().
Signed-off-by: Josef Bacik <jbacik@redhat.com> Acked-by: Nick Piggin <npiggin@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>
It's really not right to use 'access_ok()', since that is meant for the
normal "get_user()" and "copy_from/to_user()" accesses, which are done
through the TLB, rather than through the page tables.
Why? access_ok() does both too few, and too many checks. Too many,
because it is meant for regular kernel accesses that will not honor the
'user' bit in the page tables, and because it honors the USER_DS vs
KERNEL_DS distinction that we shouldn't care about in GUP. And too few,
because it doesn't do the 'canonical' check on the address on x86-64,
since the TLB will do that for us.
So instead of using a function that isn't meant for this, and does
something else and much more complicated, just do the real rules: we
don't want the range to overflow, and on x86-64, we want it to be a
canonical low address (on 32-bit, all addresses are canonical).
Acked-by: Ingo Molnar <mingo@elte.hu> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
vmscan: do not unconditionally treat zones that fail zone_reclaim() as full
On NUMA machines, the administrator can configure zone_reclaim_mode that
is a more targetted form of direct reclaim. On machines with large NUMA
distances for example, a zone_reclaim_mode defaults to 1 meaning that
clean unmapped pages will be reclaimed if the zone watermarks are not
being met. The problem is that zone_reclaim() failing at all means the
zone gets marked full.
This can cause situations where a zone is usable, but is being skipped
because it has been considered full. Take a situation where a large tmpfs
mount is occuping a large percentage of memory overall. The pages do not
get cleaned or reclaimed by zone_reclaim(), but the zone gets marked full
and the zonelist cache considers them not worth trying in the future.
This patch makes zone_reclaim() return more fine-grained information about
what occured when zone_reclaim() failued. The zone only gets marked full
if it really is unreclaimable. If it's a case that the scan did not occur
or if enough pages were not reclaimed with the limited reclaim_mode, then
the zone is simply skipped.
There is a side-effect to this patch. Currently, if zone_reclaim()
successfully reclaimed SWAP_CLUSTER_MAX, an allocation attempt would go
ahead. With this patch applied, zone watermarks are rechecked after
zone_reclaim() does some work.
This bug was introduced by commit 9276b1bc96a132f4068fdee00983c532f43d3a26
("memory page_alloc zonelist caching speedup") way back in 2.6.19 when the
zonelist_cache was introduced. It was not intended that zone_reclaim()
aggressively consider the zone to be full when it failed as full direct
reclaim can still be an option. Due to the age of the bug, it should be
considered a -stable candidate.
x86, setup (2.6.30-stable) fix 80x34 and 80x60 console modes
Note: this is not in upstream since upstream is not affected due to the
new "BIOS glovebox" subsystem.
As coded, most INT10 calls in video-vga.c allow the compiler to assume
EAX remains unchanged across them, which is not always the case. This
triggers an optimisation issue that causes vga_set_vertical_end() to be
called with an incorrect number of scanlines. Fix this by beefing up
the asm constraints on these calls.
Reported-by: Marc Aurele La France <tsi@xfree86.org> Signed-off-by: Marc Aurele La France <tsi@xfree86.org> Acked-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There are two reasons to expose the memory *a in the asm:
1) To prevent the compiler from discarding a preceeding write to *a, and
2) to prevent it from caching *a in a register over the asm.
The change has had a few days testing with a SMP build of 2.6.22.19
running on a rp3440.
This patch is about the correctness of the __ldcw() macro itself.
The use of the macro should be confined to small inline functions
to try to limit the effect of clobbering memory on GCC's optimization
of loads and stores.
The TLB flushing functions on hppa, which causes PxTLB broadcasts on the system
bus, needs to be protected by irq-safe spinlocks to avoid irq handlers to deadlock
the kernel. The deadlocks only happened during I/O intensive loads and triggered
pretty seldom, which is why this bug went so long unnoticed.
Signed-off-by: Helge Deller <deller@gmx.de>
[edited to use spin_lock_irqsave on UP as well since we'd been locking there
all this time anyway, --kyle] Signed-off-by: Kyle McMartin <kyle@mcmartin.ca> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
AMI BIOS with low memory corruption was found on Intel DG45ID
board (Bug 13710). Add this board to the blacklist - in the
(somewhat optimistic) hope of future boards/BIOSes from Intel
not having this bug.
The movq instruction, generated by __put_user_asm() when used for
64-bit data, takes a sign-extended immediate ("e") not a zero-extended
immediate ("Z").
Signed-off-by: H. Peter Anvin <hpa@zytor.com> Cc: Uros Bizjak <ubizjak@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
arch/x86/include/asm/uaccess_64.h uses wrong asm operand constraint
("ir") for movq insn. Since movq sign-extends its immediate operand,
"er" constraint should be used instead.
Attached patch changes all uses of __put_user_asm in uaccess_64.h to use
"er" when "q" insn suffix is involved.
Patch was compile tested on x86_64 with defconfig.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Timer interrupts are excluded from being disabled during suspend. The
clock events code manages the disabling of clock events on its own
because the timer interrupt needs to be functional before the resume
code reenables the device interrupts.
The mfgpt timer request its interrupt without setting the IRQF_TIMER
flag so suspend_device_irqs() disables it as well which results in a
fatal resume failure.
Adding IRQF_TIMER to the interupt flags when requesting the mrgpt
timer interrupt solves the problem.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <new-submission> Cc: Andres Salomon <dilinger@debian.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The recent commit 7513c2a761d69d2a93f17146b3563527d3618ba0 (dm raid1:
add is_remote_recovering hook for clusters) changed do_writes() to
update the ms->writes list but forgot to wake up kmirrord to process it.
The rule is that when anything is being added on ms->reads, ms->writes
or ms->failures and the list was empty before we must call
wakeup_mirrord (for immediate processing) or delayed_wake (for delayed
processing). Otherwise the bios could sit on the list indefinitely.
commit e3c8ca8336 (sched: do not count frozen tasks toward load) broke
the nr_uninterruptible accounting on freeze/thaw. On freeze the task
is excluded from accounting with a check for (task->flags &
PF_FROZEN), but that flag is cleared before the task is thawed. So
while we prevent that the task with state TASK_UNINTERRUPTIBLE
is accounted to nr_uninterruptible on freeze we decrement
nr_uninterruptible on thaw.
Use a separate flag which is handled by the freezing task itself. Set
it before calling the scheduler with TASK_UNINTERRUPTIBLE state and
clear it after we return from frozen state.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
x86: multi pci root bus with different io resource range, on
64-bit
It turns out that:
1. that AMD-based systems have two HT chains.
2. BIOS doesn't allocate resources for BAR 6 of devices under 8132 etc
3. that multi-peer-root patch will try to split root resources to peer
root resources according to PCI conf of NB
4. PCI core assigns unassigned resources, but they overlap with BARs
that are used by ioapic addr of io4 and 8132.
The reason: at that point ioapic address are not inserted yet. Solution
is to insert ioapic resources into the tree a bit earlier.
Reported-by: Stephen Frost <sfrost@snowman.net> Reported-and-Tested-by: dann frazier <dannf@hp.com> Signed-off-by: Yinghai Lu <yinghai@kernel.org> Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We want to set _spin_*:traceon:10 to set_ftrace_filter, it complains
with "Invalid argument", but the operation is successful.
This is because ftrace_process_regex() returns the number of functions that
matched the pattern. If the number is not 0, this value is returned
by ftrace_regex_write() whereas we want to return the number of bytes
virtually written.
Also the file offset pointer is not updated in this case.
If the number of matched functions is lower than the number of bytes written
by the user, this results to a reprocessing of the string given by the user with
a lower size, leading to a malformed ftrace regex and then a -EINVAL returned.
So, this patch fixes it by returning 0 if no error occured.
The fix also applies on 2.6.30
Fixes an easily triggerable BUG() when setting process affinities.
Make sure to count the number of migratable tasks in the same place:
the root rt_rq. Otherwise the number doesn't make sense and we'll hit
the BUG in set_cpus_allowed_rt().
Also, make sure we only count tasks, not groups (this is probably
already taken care of by the fact that rt_se->nr_cpus_allowed will be 0
for groups, but be more explicit)
This fixes a bug that checkpoint count gets wrong on errors when
deleting a series of checkpoints.
The count error is persistent since the checkpoint count is stored on
disk. Some userland programs refer to the count via ioctl, and this
bugfix is needed to prevent malfunction of such programs.
Leandro Lucarella gave me a report that nilfs gets stuck after its
write function fails.
The problem turned out to be caused by bugs which leave writeback flag
on pages. This fixes the problem by ensuring to clear the writeback
flag in error path.
=
The semantics of SG_DXFER_TO_FROM_DEV were:
- copy user space buffer to kernel (LLD) buffer
- do SCSI command which is assumed to be of the DATA_IN
(data from device) variety. This would overwrite
some or all of the kernel buffer
- copy kernel (LLD) buffer back to the user space.
The idea was to detect short reads by filling the original
user space buffer with some marker bytes ("0xec" it would
seem in this report). The "resid" value is a better way
of detecting short reads but that was only added this century
and requires co-operation from the LLD.
=
This patch changes the block layer mapping API to support this
semantics. This simply adds another field to struct rq_map_data and
enables __bio_copy_iov() to copy data from user space even with READ
requests.
It's better to add the flags field and kills null_mapped and the new
from_user fields in struct rq_map_data but that approach makes it
difficult to send this patch to stable trees because st and osst
drivers use struct rq_map_data (they were converted to use the block
layer in 2.6.29 and 2.6.30). Well, I should clean up the block layer
mapping API.
zhou sf reported this regiression and tested this patch:
> Since the fixmap pages are assigned higher address to lower,
> text_poke() has to use it with inverted order (FIX_TEXT_POKE1
> to FIX_TEXT_POKE0).
I prefer to just invert the order of the fixmap declaration.
It's simpler and more straightforward.
Backward fixmaps seems to be used by both x86 32 and 64.
It's really rare but a nasty bug, because it only hurts when
instructions to patch are crossing a page boundary. If this
happens, the fixmap write accesses will spill on the following
fixmap, which may very well crash the system. And this does not
crash the system, it could leave illegal instructions in place.
Thanks Masami for finding this.
It seems to have crept into the 2.6.30-rc series, so this calls
for a -stable inclusion.
The merge of the 32- and 64-bit fixmap headers made a latent
bug on x86-64 a real one: with the right config settings
it is possible for FIX_OHCI1394_BASE to overlap the FIX_BTMAP_*
range.
The buffer for the consoles are unconditionally allocated at con_init()
time, which miss the creation of the vcs(a) devices.
Since 2.6.30 (commit 4995f8ef9d3aac72745e12419d7fbaa8d01b1d81, 'vcs:
hook sysfs devices into object lifetime instead of "binding"' to be
exact) these devices are no longer created at open() and removed on
close(), but controlled by the lifetime of the buffers.
git commit f67f129e "Driver core: implement uevent suppress in kobject"
contains this chunk for fs/partitions/check.c:
/* suppress uevent if the disk supresses it */
- if (!ddev->uevent_suppress)
+ if (!dev_get_uevent_suppress(pdev))
kobject_uevent(&pdev->kobj, KOBJ_ADD);
However that should have been
- if (!ddev->uevent_suppress)
+ if (!dev_get_uevent_suppress(ddev))
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Acked-by: Ming Lei <tom.leiming@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Standard data flow for MMC/SD/SDIO cards requires that the mvsdio
controller be set for big endian operation. This is causing problems
with buffers which length is not a multiple of 4 bytes as the last
partial word doesn't get shifted all the way and stored properly in
memory. Let's compensate for this.
Gateway T6834c laptops need EAPD always on while the default behavior
for the STAC9205 reference board is to turn it off upon every HP plug.
By using the special "eapd" model, which is first introduced for Gateway
T1616 laptops for this same reason, this peculiarity can be properly
handled.
The recent rewrite of the codec parser for STAC9872 caused a regression
for some Sony VAIO models that don't give proper pin default configs
by BIOS. Even using model=vaio doesn't work because the pin definitions
are set after the pin overrides.
This patch fixes the pin definitions in patch_stac9872() to be put
in the right place before the pin overrides. Also the patch adds the
new quirk entry for VAIO F/S to have the correct pin default configs.
The capture buffer size with 64kB seems broken with CA0106.
At least, either the update timing or the DMA position is wrong,
and this screws up pulseaudio badly.
This patch restricts the max buffer size less than that to make life
a bit easier.
This patch (as1259b) makes ehci-hcd return the total number of bytes
transferred in urb->actual_length for Isochronous transfers.
Until now, the actual_length value was unaccountably left at 0.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Acked-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
commit b7af0bb ("USB: allow malformed LANGID descriptors") broke support
for devices without string descriptor support.
Reporting string descriptors is optional to USB devices, and a device
lets us know it can't deal with strings by responding to the LANGID
request with a STALL token.
The kernel handled that correctly before b7af0bb came in, but failed
hard if the LANGID was reported but broken. More than that, if a device
was not able to provide string descriptors, the LANGID was retrieved
over and over again at each string read request.
This patch changes the behaviour so that
a) the LANGID is only queried once
b) devices which can't handle string requests are not asked again
c) devices with malformed LANGID values have a sane fallback to 0x0409
Signed-off-by: Daniel Mack <daniel@caiaq.de> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The reworked Ethernet gadget has an RNDIS interop problem when used
with the CDC subset driver ... e.g. on PXA 2xx and 3xx hardware,
which currently has a hard time talking to MS-Windows hosts.
The issue is that Microsoft requires USB_CLASS_COMM. Fix by tweaking
the CDC subset driver to not switch to USB_CLASS_VENDOR_SPEC if RNDIS
is used in some other device configuration.
[ UPDATED: some "statements" were comma-terminated; fix that. ]
commit 1a1fab513734b3a4fca1bee8229e5ff7e1cb873c accidentally added the
device id to both tables in the driver, which causes problems as this is
only a single port device, not a multiple port device.
This patch (as1262) fixes a bug in usbfs: It refuses to accept
zero-length transfers, and it insists that the buffer pointer be valid
even if there is no data being transferred.
The patch also consolidates a bunch of repetitive access_ok() checks
into a single check, which incidentally fixes the lack of such a check
for Isochronous URBs.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I repeatedly get __stack_chk_fail panic()s with this driver before
applying the attached fix.
ieee80211_wx_get_name() ignores sizeof(wrqu->name) which is IFNAMSIZ (16), and
on certain conditions, the concatenated string will be larger than IFNAMSIZ
including the terminating zero.
The code in cfg80211's cfg80211_bss_update erroneously
grabs a reference to the BSS, which means that it will
never be freed.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
ISDN connection setup failed if the "connection active" and
"B channel up" messages from the device arrived in a different
order than expected. Modify the state machine to accept them in
any order.
Impact: bugfix
Signed-off-by: Tilman Schmidt <tilman@imap.cc> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The bit that tells us whether a statistics counter snapshot operation
has completed is located in the GLOBAL register block, not in the
GLOBAL2 register block, so fix up mv88e6xxx_stats_wait() to poll the
right register address.
Signed-off-by: Stephane Contri <Stephane.Contri@grassvalley.com> Signed-off-by: Lennert Buytenhek <buytenh@marvell.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Jesper noted that kmem_cache_destroy() invokes synchronize_rcu() rather than
rcu_barrier() in the SLAB_DESTROY_BY_RCU case, which could result in RCU
callbacks accessing a kmem_cache after it had been destroyed.
Acked-by: Matt Mackall <mpm@selenic.com> Reported-by: Jesper Dangaard Brouer <hawk@comx.dk> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Blue Microphones USB devices have an alternate setting that sends two
channels of data to the computer. Unfortunately, the descriptors of
that altsetting have a wrong channel setting, which means that any
recorded data from such a device has twice the sample rate from what
would be expected.
This patch adds a workaround to ignore that altsetting. Since these
devices have only one actual channel, no data is lost.
nfsd_open() gets an unrefcounted pointer to the current process's effective
credentials at the top of the function, then calls nfsd_setuser() via
fh_verify() - which may replace and destroy the current process's effective
credentials - and then passes the unrefcounted pointer to dentry_open() - but
the credentials may have been destroyed by this point.
Instead, the value from current_cred() should be passed directly to
dentry_open() as one of its arguments, rather than being cached in a variable.
Possibly fh_verify() should return the creds to use.
Signed-off-by: David Howells <dhowells@redhat.com> Tested-and-Verified-By: Steve Dickson <steved@redhat.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I recently discovered on my zalon that if the attachment fails because
of a bus misconfiguration (I scrapped my HVD array, so the card is now
unterminated) then the system oopses. The reason is that if
ncr_attach() returns NULL (signalling failure) that NULL is passed by
the goto failed straight into ncr_detach() which oopses.
The fix is just to return -ENODEV in this case.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There's some odd bug in gcc-4.2 where it miscompiles a simple loop whent
he loop counter is of type 'unsigned char' and it should count to 128.
The compiler will incorrectly decide that a trivial loop like this:
unsigned char i, ...
for (i = 0; i < 128; i++) {
..
is endless, and will compile it to a single instruction that just
branches to itself.
This was triggered by the addition of '-fno-strict-overflow', and we
could play games with compiler versions and go back to '-fwrapv'
instead, but the trivial way to avoid it is to just make the loop
induction variable be an 'int' instead.
Thanks to Krzysztof Oledzki for reporting and testing and to Troy Moure
for digging through assembler differences and finding it.
Reported-and-tested-by: Krzysztof Oledzki <olel@ans.pl> Found-by: Troy Moure <twmoure@szypr.net> Gcc-bug-acked-by: Ian Lance Taylor <iant@google.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This causes kernel images that don't run init to completion with certain
broken gcc versions.
This fixes kernel bugzilla entry:
http://bugzilla.kernel.org/show_bug.cgi?id=13012
I suspect the gcc problem is this:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28230
Fix the problem by using the -fno-strict-overflow flag instead, which
not only does not exist in the known-to-be-broken versions of gcc (it
was introduced later than fwrapv), but seems to be much less disturbing
to gcc too: the difference in the generated code by -fno-strict-overflow
are smaller (compared to using neither flag) than when using -fwrapv.
Reported-by: Barry K. Nathan <barryn@pobox.com> Pushed-by: Frans Pop <elendil@planet.nl> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
On 64 bit systems -- where sizeof(ssize_t) > sizeof(int) -- the following test
exposes a bug due to a non-careful return of an int or unsigned value:
implement a FUSE filesystem which sends an unsolicited notification to
the kernel with invalid opcode. The respective write to /dev/fuse
will return (1 << 32) - EINVAL with errno == 0 instead of -1 with
errno == EINVAL.
This fixes kernel.org bug #13584. The IOVA code attempted to optimise
the insertion of new ranges into the rbtree, with the unfortunate result
that some ranges just didn't get inserted into the tree at all. Then
those ranges would be handed out more than once, and things kind of go
downhill from there.
After discovering that we don't listen to gratuitious arps in 2.6.30
I tracked the failure down to this commit.
The patch makes absolutely no sense. RFC2131 RFC3927 and RFC5227.
are all in agreement that an arp request with sip == 0 should be used
for the probe (to prevent learning) and an arp request with sip == tip
should be used for the gratitous announcement that people can learn
from.
It appears the author of the broken patch got those two cases confused
and modified the code to drop all gratuitous arp traffic. Ouch!
Signed-off-by: Eric W. Biederman <ebiederm@aristanetworks.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When an md device is created by name (rather than number) we need to
check that the name is not already in use. If this check finds a
duplicate, we return an error without dropping the lock or freeing
the newly create mddev.
This patch fixes that.
md allows write to regions on an array to be suspended temporarily.
This allows user-space to participate is aspects of reshape.
In particular, data can be copied with not risk of a race.
We should not be blocking read requests though, so don't.
The next_ordered flag is only meaningful for devices that use __make_request.
So move the test against next_ordered out of generic code and in to
__make_request
Since this test was added, barriers have not worked on md or any
devices that don't use __make_request and so don't bother to set
next_ordered. (dm explicitly sets something other than
QUEUE_ORDERED_NONE since
commit 99360b4c18f7675b50d283301d46d755affe75fd
but notes in the comments that it is otherwise meaningless).
This patch fixes a bug in the overlap function which returned true if
one region ends exactly before the second region begins. This is no
overlap but the function returned true in that case.
alpha percpu access requires custom SHIFT_PERCPU_PTR() definition for
modules to work around addressing range limitation. This is done via
generating inline assembly using C preprocessing which forces the
assembler to generate external reference. This happens behind the
compiler's back and makes the compiler think that static percpu variables
in modules are unused.
This used to be worked around by using __unused attribute for percpu
variables which prevent the compiler from omitting the variable; however,
recent declare/definition attribute unification change broke this as
__used can't be used for declaration. Also, in the process,
PER_CPU_ATTRIBUTES definition in alpha percpu.h got broken.
This patch adds PER_CPU_DEF_ATTRIBUTES which is only used for definitions
and make alpha use it to add __used for percpu variables in modules. This
also fixes the PER_CPU_ATTRIBUTES double definition bug.
Signed-off-by: Tejun Heo <tj@kernel.org> Tested-by: maximilian attems <max@stro.at> Acked-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru> Cc: Richard Henderson <rth@twiddle.net> 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>
get_futex_key() can infinitely loop if it is called on a
virtual address that is within a huge page but not aligned to
the beginning of that page. The call to get_user_pages_fast
will return the struct page for a sub-page within the huge page
and the check for page->mapping will always fail.
The fix is to call compound_head on the page before checking
that it's mapped.
commit 64d1304a64 (futex: setup writeable mapping for futex ops which
modify user space data) did address only half of the problem of write
access faults.
The patch was made on two wrong assumptions:
1) access_ok(VERIFY_WRITE,...) would actually check write access.
On x86 it does _NOT_. It's a pure address range check.
2) a RW mapped region can not go away under us.
That's wrong as well. Nobody can prevent another thread to call
mprotect(PROT_READ) on that region where the futex resides. If that
call hits between the get_user_pages_fast() verification and the
actual write access in the atomic region we are toast again.
The solution is to not rely on access_ok and get_user() for any write
access related fault on private and shared futexes. Instead we need to
fault it in with verification of write access.
There is no generic non destructive write mechanism which would fault
the user page in trough a #PF, but as we already know that we will
fault we can as well call get_user_pages() directly and avoid the #PF
overhead.
If get_user_pages() returns -EFAULT we know that we can not fix it
anymore and need to bail out to user space.
Remove a bunch of confusing comments on this issue as well.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Commit 6b3087c6 (which introduced Blackfin SMP) broke command line passing
when the DEBUG_DOUBLEFAULT config option was enabled. Switch the code to
using a scratch register and not R7 which holds the command line.
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When a low priority interrupt (like ethernet) is triggered between 2 high
priority IPI messages, a deadlock in disable_irq() is hit by the second
IPI handler. This is because the second IPI message is queued within the
first IPI handler, but the handler doesn't process all messages, and new
ones are inserted rather than appended. So now we process all the pending
messages, and append new ones to the pending list.
URL: http://blackfin.uclinux.org/gf/tracker/5226
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com> Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
With the common IRQ code initializing much more of the irq_desc state, we
can't blindly initialize it ourselves to the local bad_irq state. If we
do, we end up wrongly clobbering many fields. So punt most of the bad irq
code as the common layers will handle the default state, and simply call
handle_bad_irq() directly when the IRQ we are processing is invalid.
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We read the SWRST (Software Reset) register to get at the last reset
state, and then we may configure the DOUBLE_FAULT bit to control behavior
when a double fault occurs. But if the lower bits of the register is
already set (like UART boot mode on a BF54x), we inadvertently make the
system reset by writing to the SYSTEM_RESET field at the same time. So
make sure the lower 4 bits are always cleared.
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com> Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We have found that the current PER_CLEAR_ON_SETID mask on Linux doesn't
include neither ADDR_COMPAT_LAYOUT, nor MMAP_PAGE_ZERO.
The current mask is READ_IMPLIES_EXEC|ADDR_NO_RANDOMIZE.
We believe it is important to add MMAP_PAGE_ZERO, because by using this
personality it is possible to have the first page mapped inside a
process running as setuid root. This could be used in those scenarios:
- Exploiting a NULL pointer dereference issue in a setuid root binary
- Bypassing the mmap_min_addr restrictions of the Linux kernel: by
running a setuid binary that would drop privileges before giving us
control back (for instance by loading a user-supplied library), we
could get the first page mapped in a process we control. By further
using mremap and mprotect on this mapping, we can then completely
bypass the mmap_min_addr restrictions.
Less importantly, we believe ADDR_COMPAT_LAYOUT should also be added
since on x86 32bits it will in practice disable most of the address
space layout randomization (only the stack will remain randomized).
Signed-off-by: Julien Tinnes <jt@cr0.org> Signed-off-by: Tavis Ormandy <taviso@sdf.lonestar.org> Acked-by: Christoph Hellwig <hch@infradead.org> Acked-by: Kees Cook <kees@ubuntu.com> Acked-by: Eugene Teo <eugene@redhat.com>
[ Shortened lines and fixed whitespace as per Christophs' suggestion ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix NULL pointer dereference in tun_chr_pool() introduced by commit 33dccbb050bbe35b88ca8cf1228dcf3e4d4b3554 ("tun: Limit amount of queued
packets per device") and triggered by this code:
This patch removes the dependency of mmap_min_addr on CONFIG_SECURITY.
It also sets a default mmap_min_addr of 4096.
mmapping of addresses below 4096 will only be possible for processes
with CAP_SYS_RAWIO.
Signed-off-by: Christoph Lameter <cl@linux-foundation.org> Acked-by: Eric Paris <eparis@redhat.com> Looks-ok-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: James Morris <jmorris@namei.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Turning on this flag could prevent the compiler from optimising away
some "useless" checks for null pointers. Such bugs can sometimes become
exploitable at compile time because of the -O2 optimisation.
See http://gcc.gnu.org/onlinedocs/gcc-4.1.2/gcc/Optimize-Options.html
An example that clearly shows this 'problem' is commit 6bf67672.
The file opened in acct_on and freshly stored in the ns->bacct struct can
be closed in acct_file_reopen by a concurrent call after we release
acct_lock and before we call mntput(file->f_path.mnt).
Record file->f_path.mnt in a local variable and use this variable only.
Signed-off-by: Renaud Lottiaux <renaud.lottiaux@kerlabs.com> Signed-off-by: Louis Rilling <louis.rilling@kerlabs.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>
As it turns out, the call trace is messed up due to gcc's inlining, but
I isolated the problem anyway: kvm_write_guest_time() is being used in a
non-thread-safe manner on preemptable kernels.
Basically kvm_write_guest_time()'s body needs to be surrounded by
preempt_disable() and preempt_enable(), since the kernel won't let us
query any per-CPU data (indirectly using smp_processor_id()) without
preemption disabled. The attached patch fixes this issue by disabling
preemption inside kvm_write_guest_time().
[marcelo: surround only __get_cpu_var calls since the warning
is harmless]
We need to save register state *after* idling GEM, clearing the ring,
and uninstalling the IRQ handler, or we might end up saving bogus
fence regs, for one. Our restore ordering should already be correct,
since we do GEM, ring and IRQ init after restoring the last register
state, which prevents us from clobbering things.
I put this together to potentially address a bug, but I haven't heard
back if it fixes it yet. However I think it stands on its own, so I'm
sending it in.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org> Signed-off-by: Eric Anholt <eric@anholt.net> Cc: Jie Luo <clotho67@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
With 2.6.30, the error handling code in cdrom_newpc_intr was changed
to deal with partial request failures by normally completing the 'good'
parts of a request and only 'error' the last (and presumably,
incompletely transferred) bio associated with a particular
request. In order to do this, ide_complete_rq is called over
ide_cd_error_cmd() to partially complete the rq. The block layer
does partial completion only for requests with bio's and if the
rq doesn't have one (eg 'GPCMD_READ_DISC_INFO') the request is
completed as a whole and the drive->hwif->rq pointer set to NULL
afterwards. When calling ide_complete_rq again to report
the error, this null pointer is derefenced, resulting in a kernel
crash.
This fixes http://bugzilla.kernel.org/show_bug.cgi?id=13399.
Since early printk only makes sense/works when the serial driver is built
into the kernel, disable the option for this driver when it is going to be
built as a module. Otherwise we get build failures due to the ifdef
handling.
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Alan Cox <alan@linux.intel.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There is a call to write_lock() in gen_pool_destroy which is not balanced
by any corresponding write_unlock(). This causes problems with preemption
because the preemption-disable counter is incremented in the write_lock()
call, but never decremented by any call to write_unlock(). This bug is
gen_pool_destroy, and one of them is non-x86 arch-specific code.
On NUMA machines, the administrator can configure zone_reclaim_mode that
is a more targetted form of direct reclaim. On machines with large NUMA
distances for example, a zone_reclaim_mode defaults to 1 meaning that
clean unmapped pages will be reclaimed if the zone watermarks are not
being met.
There is a heuristic that determines if the scan is worthwhile but it is
possible that the heuristic will fail and the CPU gets tied up scanning
uselessly. Detecting the situation requires some guesswork and
experimentation so this patch adds a counter "zreclaim_failed" to
/proc/vmstat. If during high CPU utilisation this counter is increasing
rapidly, then the resolution to the problem may be to set
/proc/sys/vm/zone_reclaim_mode to 0.
[akpm@linux-foundation.org: name things consistently] Signed-off-by: Mel Gorman <mel@csn.ul.ie> Reviewed-by: Rik van Riel <riel@redhat.com> Cc: Christoph Lameter <cl@linux-foundation.org> Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Wu Fengguang <fengguang.wu@intel.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>
A bug was brought to my attention against a distro kernel but it affects
mainline and I believe problems like this have been reported in various
guises on the mailing lists although I don't have specific examples at the
moment.
The reported problem was that malloc() stalled for a long time (minutes in
some cases) if a large tmpfs mount was occupying a large percentage of
memory overall. The pages did not get cleaned or reclaimed by
zone_reclaim() because the zone_reclaim_mode was unsuitable, but the lists
are uselessly scanned frequencly making the CPU spin at near 100%.
This patchset intends to address that bug and bring the behaviour of
zone_reclaim() more in line with expectations which were noticed during
investigation. It is based on top of mmotm and takes advantage of
Kosaki's work with respect to zone_reclaim().
Patch 1 fixes the heuristics that zone_reclaim() uses to determine if the
scan should go ahead. The broken heuristic is what was causing the
malloc() stall as it uselessly scanned the LRU constantly. Currently,
zone_reclaim is assuming zone_reclaim_mode is 1 and historically it
could not deal with tmpfs pages at all. This fixes up the heuristic so
that an unnecessary scan is more likely to be correctly avoided.
Patch 2 notes that zone_reclaim() returning a failure automatically means
the zone is marked full. This is not always true. It could have
failed because the GFP mask or zone_reclaim_mode were unsuitable.
Patch 3 introduces a counter zreclaim_failed that will increment each
time the zone_reclaim scan-avoidance heuristics fail. If that
counter is rapidly increasing, then zone_reclaim_mode should be
set to 0 as a temporarily resolution and a bug reported because
the scan-avoidance heuristic is still broken.
This patch:
On NUMA machines, the administrator can configure zone_reclaim_mode that
is a more targetted form of direct reclaim. On machines with large NUMA
distances for example, a zone_reclaim_mode defaults to 1 meaning that
clean unmapped pages will be reclaimed if the zone watermarks are not
being met.
There is a heuristic that determines if the scan is worthwhile but the
problem is that the heuristic is not being properly applied and is
basically assuming zone_reclaim_mode is 1 if it is enabled. The lack of
proper detection can manfiest as high CPU usage as the LRU list is scanned
uselessly.
Historically, once enabled it was depending on NR_FILE_PAGES which may
include swapcache pages that the reclaim_mode cannot deal with. Patch
vmscan-change-the-number-of-the-unmapped-files-in-zone-reclaim.patch by
Kosaki Motohiro noted that zone_page_state(zone, NR_FILE_PAGES) included
pages that were not file-backed such as swapcache and made a calculation
based on the inactive, active and mapped files. This is far superior when
zone_reclaim==1 but if RECLAIM_SWAP is set, then NR_FILE_PAGES is a
reasonable starting figure.
This patch alters how zone_reclaim() works out how many pages it might be
able to reclaim given the current reclaim_mode. If RECLAIM_SWAP is set in
the reclaim_mode it will either consider NR_FILE_PAGES as potential
candidates or else use NR_{IN}ACTIVE}_PAGES-NR_FILE_MAPPED to discount
swapcache and other non-file-backed pages. If RECLAIM_WRITE is not set,
then NR_FILE_DIRTY number of pages are not candidates. If RECLAIM_SWAP is
not set, then NR_FILE_MAPPED are not.
[kosaki.motohiro@jp.fujitsu.com: Estimate unmapped pages minus tmpfs pages]
[fengguang.wu@intel.com: Fix underflow problem in Kosaki's estimate] Signed-off-by: Mel Gorman <mel@csn.ul.ie> Reviewed-by: Rik van Riel <riel@redhat.com> Acked-by: Christoph Lameter <cl@linux-foundation.org> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Cc: Wu Fengguang <fengguang.wu@intel.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>
If someone changes the size of the device simultaneously, i_size_read
is guaranteed to return a valid value (either the old one or the new one).
i_size can return some intermediate invalid value (on 32-bit computers
with 64-bit i_size, the reads to both halves of i_size can be interleaved
with updates to i_size, resulting in garbage being returned).
Cc: Yi Yang <yi.y.yang@intel.com> Cc: Jonathan Brassow <jbrassow@redhat.com> Cc: stable@kernel.org Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Milan Broz <mbroz@redhat.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When snapshots are created using 'p' instead of 'P' as the
exception store type, the device-mapper table loading fails.
This patch makes the code case insensitive as intended and fixes some
regressions reported with device-mapper snapshots.
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: Arjan van de Ven <arjan@infradead.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There is a possibility of race condition because kevent queue is not flushed
in the multipath destructor. The scenario is:
- some event happens and is queued to keventd
- keventd thread is delayed due to scheuling latency or some other work
- multipath device is destroyed
- keventd now attempts to process work_struct that is residing in already
released memory.
The patch flushes the keventd queue in multipath constructor.
I've already fixed similar bug in dm-raid1.
projects / karo-tx-linux.git / log