Dmitry Vyukov reported a divide by 0 triggered by syzkaller, exploiting
tcp_disconnect() path that was never really considered and/or used
before syzkaller ;)
I was not able to reproduce the bug, but it seems issues here are the
three possible actions that assumed they would never trigger on a
listener.
1) tcp_write_timer_handler
2) tcp_delack_timer_handler
3) MTU reduction
Only IPv6 MTU reduction was properly testing TCP_CLOSE and TCP_LISTEN
states from tcp_v6_mtu_reduced()
Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: Dmitry Vyukov <dvyukov@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The code where sk_clone() came from created a new socket and locked it,
but then, on the error path didn't unlock it.
This problem stayed there for a long while, till b0691c8ee7c2 ("net:
Unlock sock before calling sk_free()") fixed it, but unfortunately the
callers of sk_clone() (now sk_clone_locked()) were not audited and the
one in dccp_create_openreq_child() remained.
Now in the age of the syskaller fuzzer, this was finally uncovered, as
reported by Dmitry:
---- 8< ----
I've got the following report while running syzkaller fuzzer on 86292b33d4b7 ("Merge branch 'akpm' (patches from Andrew)")
It is now very clear that silly TCP listeners might play with
enabling/disabling timestamping while new children are added
to their accept queue.
Meaning net_enable_timestamp() can be called from BH context
while current state of the static key is not enabled.
Lets play safe and allow all contexts.
The work queue is scheduled only under the problematic cases,
which are the static key enable/disable transition, to not slow down
critical paths.
This extends and improves what we did in commit 5fa8bbda38c6 ("net: use
a work queue to defer net_disable_timestamp() work")
Fixes: b90e5794c5bd ("net: dont call jump_label_dec from irq context") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: Dmitry Vyukov <dvyukov@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
This happens because addr.sa_data copied from the userspace is not
zero-terminated, and copying it with strlcpy() in packet_bind_spkt()
results in calling strlen() on the kernel copy of that non-terminated
buffer.
Signed-off-by: Alexander Potapenko <glider@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
l2tp_ip_backlog_recv may not return -1 if the packet gets dropped.
The return value is passed up to ip_local_deliver_finish, which treats
negative values as an IP protocol number for resubmission.
Signed-off-by: Paul Hüber <phueber@kernsp.in> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Fix a smatch warning:
drivers/scsi/mvsas/mv_sas.c:740 mvs_task_prep() warn: curly braces intended?
The code is correct, the indention is misleading. When the device is not
ready we want to return SAS_PHY_DOWN. But current indentation makes it
look like we only do so in the else branch of if (mvi_dev).
Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com> Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Fix a NULL-pointer dereference in the interrupt callback should a
malicious device send data containing a bad port number by adding the
missing sanity check.
Make sure to verify that we have the required interrupt-out endpoint for
IOWarrior56 devices to avoid dereferencing a NULL-pointer in write
should a malicious device lack such an endpoint.
Fixes: 946b960d13c1 ("USB: add driver for iowarrior devices.") Signed-off-by: Johan Hovold <johan@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Make sure to check for the required interrupt-in endpoint to avoid
dereferencing a NULL-pointer should a malicious device lack such an
endpoint.
Note that a fairly recent change purported to fix this issue, but added
an insufficient test on the number of endpoints only, a test which can
now be removed.
Fixes: 4ec0ef3a8212 ("USB: iowarrior: fix oops with malicious USB descriptors") Fixes: 946b960d13c1 ("USB: add driver for iowarrior devices.") Signed-off-by: Johan Hovold <johan@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
This driver needlessly took another reference to the tty on open, a
reference which was then never released on close. This lead to not just
a leak of the tty, but also a driver reference leak that prevented the
driver from being unloaded after a port had once been opened.
Upstream commit 98d74f9ceaef ("xhci: fix 10 second timeout on removal of
PCI hotpluggable xhci controllers") fixes a problem with hot pluggable PCI
xhci controllers which can result in excessive timeouts, to the point where
the system reports a deadlock.
The same problem is seen with hot pluggable xhci controllers using the
xhci-plat driver, such as the driver used for Type-C ports on rk3399.
Similar to hot-pluggable PCI controllers, the driver for this chip
removes the xhci controller from the system when the Type-C cable is
disconnected.
The solution for PCI devices works just as well for non-PCI devices
and avoids the problem.
Some gadget drivers are bad, bad boys. We notice
that ADB was passing bad Burst Size which caused top
bits of param0 to be overwritten which confused DWC3
when running this command.
In order to avoid future issues, we're going to make
sure values passed by macros are always safe for the
controller. Note that ADB still needs a fix to *not*
pass bad values.
Reported-by: Mohamed Abbas <mohamed.abbas@intel.com> Sugested-by: Adam Andruszak <adam.andruszak@intel.com> Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
There are several trace include files that define TRACE_INCLUDE_FILE.
Include several of them in the same .c file (as I currently have in
some code I am working on), and the compile will blow up with a
"warning: "TRACE_INCLUDE_FILE" redefined #define TRACE_INCLUDE_FILE syscalls"
Every other include file in include/trace/events/ avoids that issue
by having a #undef TRACE_INCLUDE_FILE before the #define; syscalls.h
should have one, too.
Link: http://lkml.kernel.org/r/20160928225554.13bd7ac6@annuminas.surriel.com Fixes: b8007ef74222 ("tracing: Separate raw syscall from syscall tracer") Signed-off-by: Rik van Riel <riel@redhat.com> Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
When expanding the la or dla pseudo-instruction in a delay slot the GNU
assembler will complain should the pseudo-instruction expand to multiple
actual instructions, since only the first of them will be in the delay
slot leading to the pseudo-instruction being only partially executed if
the branch is taken. Use of PTR_LA in the dec int-handler.S leads to
such warnings:
arch/mips/dec/int-handler.S: Assembler messages:
arch/mips/dec/int-handler.S:149: Warning: macro instruction expanded into multiple instructions in a branch delay slot
arch/mips/dec/int-handler.S:198: Warning: macro instruction expanded into multiple instructions in a branch delay slot
The #warning was present 10 years ago when the driver first got merged.
As the platform is rather obsolete by now, it seems very unlikely that
the warning will cause anyone to fix the code properly.
kernelci.org reports the warning for every build in the meantime, so
I think it's better to just turn it into a code comment to reduce
noise.
Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Over the years the code has been changed various times leading to
argc/argv being defined in a different function to where we actually
use the variables. Clean this up by moving them to prom_init_cmdline().
kernelci.org reports a warning for this driver, as it copies a local
variable into a 'const char *' string:
drivers/mtd/maps/pmcmsp-flash.c:149:30: warning: passing argument 1 of 'strncpy' discards 'const' qualifier from pointer target type [-Wdiscarded-qualifiers]
Using kstrndup() simplifies the code and avoids the warning.
Signed-off-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Marek Vasut <marek.vasut@gmail.com> Signed-off-by: Brian Norris <computersforpeace@gmail.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
kernelci reports a failure of the ip28_defconfig build after upgrading its
gcc version:
arch/mips/sgi-ip22/Platform:29: *** gcc doesn't support needed option -mr10k-cache-barrier=store. Stop.
The problem apparently is that the -mr10k-cache-barrier=store option is now
rejected for CPUs other than r10k. Explicitly including the CPU in the
check fixes this and is safe because both options were introduced in
gcc-4.4.
One of the last remaining failures in kernelci.org is for a gcc bug:
drivers/net/ethernet/qlogic/qlge/qlge_main.c:4819:1: error: insn does not satisfy its constraints:
drivers/net/ethernet/qlogic/qlge/qlge_main.c:4819:1: internal compiler error: in extract_constrain_insn, at recog.c:2190
This is apparently broken in gcc-6 but fixed in gcc-7, and I cannot
reproduce the problem here. However, it is clear that ip27_defconfig
does not actually need this driver as the platform has only PCI-X but
not PCIe, and the qlge adapter in turn is PCIe-only.
The driver was originally enabled in 2010 along with lots of other
drivers.
An ancient gcc bug (first reported in 2003) has apparently resurfaced
on MIPS, where kernelci.org reports an overly large stack frame in the
whirlpool hash algorithm:
crypto/wp512.c:987:1: warning: the frame size of 1112 bytes is larger than 1024 bytes [-Wframe-larger-than=]
With some testing in different configurations, I'm seeing large
variations in stack frames size up to 1500 bytes for what should have
around 300 bytes at most. I also checked the reference implementation,
which is essentially the same code but also comes with some test and
benchmarking infrastructure.
It seems that recent compiler versions on at least arm, arm64 and powerpc
have a partial fix for this problem, but enabling "-fsched-pressure", but
even with that fix they suffer from the issue to a certain degree. Some
testing on arm64 shows that the time needed to hash a given amount of
data is roughly proportional to the stack frame size here, which makes
sense given that the wp512 implementation is doing lots of loads for
table lookups, and the problem with the overly large stack is a result
of doing a lot more loads and stores for spilled registers (as seen from
inspecting the object code).
Disabling -fschedule-insns consistently fixes the problem for wp512,
in my collection of cross-compilers, the results are consistently better
or identical when comparing the stack sizes in this function, though
some architectures (notable x86) have schedule-insns disabled by
default.
The four columns are:
default: -O2
press: -O2 -fsched-pressure
nopress: -O2 -fschedule-insns -fno-sched-pressure
nosched: -O2 -no-schedule-insns (disables sched-pressure)
Trying the same test for serpent-generic, the picture is a bit different,
and while -fno-schedule-insns is generally better here than the default,
-fsched-pressure wins overall, so I picked that instead.
I did not do any runtime tests with serpent, so it is possible that stack
frame size does not directly correlate with runtime performance here and
it actually makes things worse, but it's more likely to help here, and
the reduced stack frame size is probably enough reason to apply the patch,
especially given that the crypto code is often used in deep call chains.
A recent change claimed to fix an off-by-one error in the OOB-port
completion handler, but instead introduced such an error. This could
specifically led to modem-status changes going unnoticed, effectively
breaking TIOCMGET.
Note that the offending commit fixes a loop-condition underflow and is
marked for stable, but should not be backported without this fix.
Reported-by: Ben Hutchings <ben@decadent.org.uk> Fixes: 2d380889215f ("USB: serial: digi_acceleport: fix OOB data sanity
check") Signed-off-by: Johan Hovold <johan@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
PCI hotpluggable xhci controllers such as some Alpine Ridge solutions will
remove the xhci controller from the PCI bus when the last USB device is
disconnected.
Add a flag to indicate that the host is being removed to avoid queueing
configure_endpoint commands for the dropped endpoints.
For PCI hotplugged controllers this will prevent 5 second command timeouts
For static xhci controllers the configure_endpoint command is not needed
in the removal case as everything will be returned, freed, and the
controller is reset.
For now the flag is only set for PCI connected host controllers.
In commit 19be0eaffa3a ("mm: remove gup_flags FOLL_WRITE games from
__get_user_pages()"), the mm code was changed from unsetting FOLL_WRITE
after a COW was resolved to setting the (newly introduced) FOLL_COW
instead. Simultaneously, the check in gup.c was updated to still allow
writes with FOLL_FORCE set if FOLL_COW had also been set.
However, a similar check in huge_memory.c was forgotten. As a result,
remote memory writes to ro regions of memory backed by transparent huge
pages cause an infinite loop in the kernel (handle_mm_fault sets
FOLL_COW and returns 0 causing a retry, but follow_trans_huge_pmd bails
out immidiately because `(flags & FOLL_WRITE) && !pmd_write(*pmd)` is
true.
While in this state the process is stil SIGKILLable, but little else
works (e.g. no ptrace attach, no other signals). This is easily
reproduced with the following code (assuming thp are set to always):
Fix this by updating follow_trans_huge_pmd in huge_memory.c analogously
to the update in gup.c in the original commit. The same pattern exists
in follow_devmap_pmd. However, we should not be able to reach that
check with FOLL_COW set, so add WARN_ONCE to make sure we notice if we
ever do.
[akpm@linux-foundation.org: coding-style fixes] Link: http://lkml.kernel.org/r/20170106015025.GA38411@juliacomputing.com Signed-off-by: Keno Fischer <keno@juliacomputing.com> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Greg Thelen <gthelen@google.com> Cc: Nicholas Piggin <npiggin@gmail.com> Cc: Willy Tarreau <w@1wt.eu> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Kees Cook <keescook@chromium.org> Cc: Andy Lutomirski <luto@kernel.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Hugh Dickins <hughd@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[bwh: Backported to 3.2:
- Drop change to follow_devmap_pmd()
- pmd_dirty() is not available; check the page flags as in
can_follow_write_pte()
- Adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
[mhocko:
This has been forward ported from the 3.2 stable tree.
And fixed to return NULL.] Reviewed-by: Michal Hocko <mhocko@suse.cz> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The total field from struct xfs_alloc_arg is a bit of an unknown
commodity. It is documented as the total block requirement for the
transaction and is used in this manner from most call sites by virtue of
passing the total block reservation of the transaction associated with
an allocation. Several xfs_bmapi_write() callers pass hardcoded values
of 0 or 1 for the total block requirement, which is a historical oddity
without any clear reasoning.
The xfs_iomap_write_direct() caller, for example, passes 0 for the total
block requirement. This has been determined to cause problems in the
form of ABBA deadlocks of AGF buffers due to incorrect AG selection in
the block allocator. Specifically, the xfs_alloc_space_available()
function incorrectly selects an AG that doesn't actually have sufficient
space for the allocation. This occurs because the args.total field is 0
and thus the remaining free space check on the AG doesn't actually
consider the size of the allocation request. This locks the AGF buffer,
the allocation attempt proceeds and ultimately fails (in
xfs_alloc_fix_minleft()), and xfs_alloc_vexent() moves on to the next
AG. In turn, this can lead to incorrect AG locking order (if the
allocator wraps around, attempting to lock AG 0 after acquiring AG N)
and thus deadlock if racing with another operation. This problem has
been reproduced via generic/299 on smallish (1GB) ramdisk test devices.
To avoid this problem, replace the undocumented hardcoded total
parameters from the iomap and utility callers to pass the block
reservation used for the associated transaction. This is consistent with
other xfs_bmapi_write() callers throughout XFS. The assumption is that
the total field allows the selection of an AG that can handle the entire
operation rather than simply the allocation/range being requested (e.g.,
resulting btree splits, etc.). This addresses the aforementioned
generic/299 hang by ensuring AG selection only occurs when the
allocation can be satisfied by the AG.
[nb] backport to 3.12
Reported-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com> Signed-off-by: Nikolay Borisov <nborisov@suse.com> Acked-by: Brian Foster <bfoster@redhat.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Commit df2cb6daa4 ("block: Avoid deadlocks with bio allocation by
stacking drivers") created a workqueue for every bio set and code
in bio_alloc_bioset() that tries to resolve some low-memory deadlocks
by redirecting bios queued on current->bio_list to the workqueue if the
system is low on memory. However other deadlocks (see below **) may
happen, without any low memory condition, because generic_make_request
is queuing bios to current->bio_list (rather than submitting them).
** the related dm-snapshot deadlock is detailed here:
https://www.redhat.com/archives/dm-devel/2016-July/msg00065.html
Fix this deadlock by redirecting any bios on current->bio_list to the
bio_set's rescue workqueue on every schedule() call. Consequently,
when the process blocks on a mutex, the bios queued on
current->bio_list are dispatched to independent workqueus and they can
complete without waiting for the mutex to be available.
The structure blk_plug contains an entry cb_list and this list can contain
arbitrary callback functions that are called when the process blocks.
To implement this fix DM (ab)uses the onstack plug's cb_list interface
to get its flush_current_bio_list() called at schedule() time.
This fixes the snapshot deadlock - if the map method blocks,
flush_current_bio_list() will be called and it redirects bios waiting
on current->bio_list to appropriate workqueues.
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1267650
Depends-on: df2cb6daa4 ("block: Avoid deadlocks with bio allocation by stacking drivers") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Recently fallocate patch was merged and it uses
MSDOS_I(inode)->mmu_private at fat_evict_inode(). However,
fat_inode/fsinfo_inode that was introduced in past didn't initialize
MSDOS_I(inode) properly.
With those combinations, it became the cause of accessing random entry
in FAT area.
The current caching state may not be tt_cached, even though the
placement contains TTM_PL_FLAG_CACHED, because placement can contain
multiple caching flags. Trying to swap out such a BO would trip up the
BUG_ON(ttm->caching_state != tt_cached);
in ttm_tt_swapout.
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com> Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com> Reviewed-by: Christian König <christian.koenig@amd.com>. Reviewed-by: Sinclair Yeh <syeh@vmware.com> Signed-off-by: Christian König <christian.koenig@amd.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The test to see if VGA was already enabled is doing an unnecessary
second test from a register that may or may not have been initialized
to a valid value. Remove it.
Signed-off-by: Y.C. Chen <yc_chen@aspeedtech.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Acked-by: Joel Stanley <joel@jms.id.au> Tested-by: Y.C. Chen <yc_chen@aspeedtech.com> Signed-off-by: Dave Airlie <airlied@redhat.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The issue was found when entering suspend and resume.
It triggers a warning in:
mac80211/key.c: ieee80211_enable_keys()
...
WARN_ON_ONCE(sdata->crypto_tx_tailroom_needed_cnt ||
sdata->crypto_tx_tailroom_pending_dec);
...
It points out sdata->crypto_tx_tailroom_pending_dec isn't cleaned up successfully
in a delayed_work during suspend. Add a flush_delayed_work to fix it.
Signed-off-by: Matt Chen <matt.chen@intel.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
FDT tag parsing is not related to whether BLK_DEV_INITRD is configured
or not, move it out of the corresponding #ifdef/#endif block.
This fixes passing external FDT to the kernel configured w/o
BLK_DEV_INITRD support.
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
It is not sufficient to just check that the lock pids match when
granting a callback, we also need to ensure that we're granting
the callback on the right file.
When calling set_mode from sys/fs, the call flow locks the sys/fs lock
first and then tries to lock rtnl_lock (when calling ipoib_set_mod).
On the other hand, the rmmod call flow takes the rtnl_lock first
(when calling unregister_netdev) and then tries to take the sys/fs
lock. Deadlock a->b, b->a.
The problem starts when ipoib_set_mod frees it's rtnl_lck and tries
to get it after that.
Return a sensible value if TASK_SIZE if called from a kernel thread.
This gets us around an issue with copy_mount_options that does a magic
size calculation "TASK_SIZE - (unsigned long)data" while in a kernel
thread and data pointing to kernel space.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
For devices with multiple input queues, tiqdio_call_inq_handlers()
iterates over all input queues and clears the device's DSCI
during each iteration. If the DSCI is re-armed during one
of the later iterations, we therefore do not scan the previous
queues again.
The re-arming also raises a new adapter interrupt. But its
handler does not trigger a rescan for the device, as the DSCI
has already been erroneously cleared.
This can result in queue stalls on devices with multiple
input queues.
Fix it by clearing the DSCI just once, prior to scanning the queues.
As the code is moved in front of the loop, we also need to access
the DSCI directly (ie irq->dsci) instead of going via each queue's
parent pointer to the same irq. This is not a functional change,
and a follow-up patch will clean up the other users.
In practice, this bug only affects CQ-enabled HiperSockets devices,
ie. devices with sysfs-attribute "hsuid" set. Setting a hsuid is
needed for AF_IUCV socket applications that use HiperSockets
communication.
Fixes: 104ea556ee7f ("qdio: support asynchronous delivery of storage blocks") Reviewed-by: Ursula Braun <ubraun@linux.vnet.ibm.com> Signed-off-by: Julian Wiedmann <jwi@linux.vnet.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The MKS Instruments SCOM-0800 and SCOM-0801 cards (originally by Tenta
Technologies) are 3U CompactPCI serial cards with 4 and 8 serial ports,
respectively. The first 4 ports are implemented by an OX16PCI954 chip,
and the second 4 ports are implemented by an OX16C954 chip on a local
bus, bridged by the second PCI function of the OX16PCI954. The ports
are jumper-selectable as RS-232 and RS-422/485, and the UARTs use a
non-standard oscillator frequency of 20 MHz (base_baud = 1250000).
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Currently N_HDLC line discipline uses a self-made singly linked list for
data buffers and has n_hdlc.tbuf pointer for buffer retransmitting after
an error.
The commit be10eb7589337e5defbe214dae038a53dd21add8
("tty: n_hdlc add buffer flushing") introduced racy access to n_hdlc.tbuf.
After tx error concurrent flush_tx_queue() and n_hdlc_send_frames() can put
one data buffer to tx_free_buf_list twice. That causes double free in
n_hdlc_release().
Let's use standard kernel linked list and get rid of n_hdlc.tbuf:
in case of tx error put current data buffer after the head of tx_buf_list.
Signed-off-by: Alexander Popov <alex.popov@linux.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The class of 4 n_hdls buf locks is the same because a single function
n_hdlc_buf_list_init is used to init all the locks. But since
flush_tx_queue takes n_hdlc->tx_buf_list.spinlock and then calls
n_hdlc_buf_put which takes n_hdlc->tx_free_buf_list.spinlock, lockdep
emits a warning:
=============================================
[ INFO: possible recursive locking detected ]
4.3.0-25.g91e30a7-default #1 Not tainted
---------------------------------------------
a.out/1248 is trying to acquire lock:
(&(&list->spinlock)->rlock){......}, at: [<ffffffffa01fd020>] n_hdlc_buf_put+0x20/0x60 [n_hdlc]
but task is already holding lock:
(&(&list->spinlock)->rlock){......}, at: [<ffffffffa01fdc07>] n_hdlc_tty_ioctl+0x127/0x1d0 [n_hdlc]
other info that might help us debug this:
Possible unsafe locking scenario:
The driver was calculating the adapter command pagesize indicator from
the system pagesize. However, the buffers the driver allocates are only
one size (SLI4_PAGE_SIZE), so no calculation was necessary.
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Cc: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
CC arch/mips/mm/sc-ip22.o
{standard input}: Assembler messages:
{standard input}:132: Error: number (0x9000000080000000) larger than 32 bits
{standard input}:159: Error: number (0x9000000080000000) larger than 32 bits
{standard input}:200: Error: number (0x9000000080000000) larger than 32 bits
scripts/Makefile.build:293: recipe for target 'arch/mips/mm/sc-ip22.o' failed
make[1]: *** [arch/mips/mm/sc-ip22.o] Error 1
MIPS has used .set mips3 to temporarily switch the assembler to 64 bit
mode in 64 bit kernels virtually forever. Binutils 2.25 broke this
behavious partially by happily accepting 64 bit instructions in .set mips3
mode but puking on 64 bit constants when generating 32 bit ELF. Binutils
2.26 restored the old behaviour again.
Fix build with binutils 2.25 by open coding the offending
dli $1, 0x9000000080000000
as
li $1, 0x9000
dsll $1, $1, 48
which is ugly be the only thing that will build on all binutils vintages.
Whenever there is a watchpoint match occurs, hw_breakpoint_handler will
be called by do_break via notifier chains mechanism. If watchpoint is
registered by xmon, hw_breakpoint_handler won't find any associated
perf_event and returns immediately with NOTIFY_STOP. Similarly, do_break
also returns without notifying to xmon.
Solve this by returning NOTIFY_DONE when hw_breakpoint_handler does not
find any perf_event associated with matched watchpoint, rather than
NOTIFY_STOP, which tells the core code to continue calling the other
breakpoint handlers including the xmon one.
Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
We're not taking into account that the space needed for the (variable
length) attr bitmap, with the result that we'd sometimes get a spurious
ERANGE when the ACL data got close to the end of a page.
Just add in an extra page to make sure.
Signed-off-by: Weston Andros Adamson <dros@primarydata.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Bitmap and attrlen follow immediately after the op reply header. This
was an oversight from commit bf118a342f.
Consequences of this are just minor efficiency (extra calls to
xdr_shrink_bufhead).
Fixes: bf118a342f10 "NFSv4: include bitmap in nfsv4 get acl data" Reviewed-by: Kinglong Mee <kinglongmee@gmail.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The RDMA core uses ib_pack() to convert from unpacked CPU structs
to on-the-wire bitpacked structs.
This process requires that 1 bit fields are declared as u8 in the
unpacked struct, otherwise the packing process does not read the
value properly and the packed result is wired to 0. Several
places wrongly used int.
Crucially this means the kernel has never, set reversible
correctly in the path record request. It has always asked for
irreversible paths even if the ULP requests otherwise.
When the kernel is used with a SM that supports this feature, it
completely breaks communication management if reversible paths are
not properly requested.
The only reason this ever worked is because opensm ignores the
reversible bit.
It may happen that not all CPUs are online when we do hv_synic_alloc() and
in case more CPUs come online later we may try accessing these allocated
structures.
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com> Signed-off-by: K. Y. Srinivasan <kys@microsoft.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
fuse_file_put() was missing the "force" flag for the RELEASE request when
sending synchronously (fuseblk).
If this flag is not set, then a sync request may be interrupted before it
is dequeued by the userspace filesystem. In this case the OPEN won't be
balanced with a RELEASE.
[js] force is a variable, not a bit
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com> Fixes: 5a18ec176c93 ("fuse: fix hang of single threaded fuseblk filesystem") Signed-off-by: Jiri Slaby <jslaby@suse.cz>
This patch fixes the OTP register definitions for the AR934x and AR9550
WMAC SoC.
Previously, the ath9k driver was unable to initialize the integrated
WMAC on an Aerohive AP121:
| ath: phy0: timeout (1000 us) on reg 0x30018: 0xbadc0ffe & 0x00000007 != 0x00000004
| ath: phy0: timeout (1000 us) on reg 0x30018: 0xbadc0ffe & 0x00000007 != 0x00000004
| ath: phy0: Unable to initialize hardware; initialization status: -5
| ath9k ar934x_wmac: failed to initialize device
| ath9k: probe of ar934x_wmac failed with error -5
It turns out that the AR9300_OTP_STATUS and AR9300_OTP_DATA
definitions contain a typo.
Cc: Gabor Juhos <juhosg@openwrt.org> Fixes: add295a4afbdf5852d0 "ath9k: use correct OTP register offsets for AR9550" Signed-off-by: Christian Lamparter <chunkeey@googlemail.com> Signed-off-by: Chris Blake <chrisrblake93@gmail.com> Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
If the file system requires journal recovery, and the device is
read-ony, return EROFS to the mount system call. This allows xfstests
generic/050 to pass.
If the journal is aborted, the needs_recovery feature flag should not
be removed. Otherwise, it's the journal might not get replayed and
this could lead to more data getting lost.
If filesystem groups are artifically small (using parameter -g to
mkfs.ext4), ext4_mb_normalize_request() can result in a request that is
larger than a block group. Trim the request size to not confuse
allocation code.
Reported-by: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
If the journal has been aborted, we shouldn't mark the underlying
buffer head as dirty, since that will cause the metadata block to get
modified. And if the journal has been aborted, we shouldn't allow
this since it will almost certainly lead to a corrupted file system.
sd_check_events() is called asynchronously, and might race
with device removal. So always take a disk reference when
processing the event to avoid the device being removed while
the event is processed.
Signed-off-by: Hannes Reinecke <hare@suse.de> Reviewed-by: Ewan D. Milne <emilne@redhat.com> Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Cc: Jinpu Wang <jinpu.wang@profitbricks.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The driver currently checks the SELF_TEST_FAILED first and then
KERNEL_PANIC next. Under error conditions(boot code failure) both
SELF_TEST_FAILED and KERNEL_PANIC can be set at the same time.
The driver has the capability to reset the controller on an KERNEL_PANIC,
but not on SELF_TEST_FAILED.
Fixed by first checking KERNEL_PANIC and then the others.
Fixes: e8b12f0fb835223752 ([SCSI] aacraid: Add new code for PMC-Sierra's SRC base controller family) Signed-off-by: Raghava Aditya Renukunta <RaghavaAditya.Renukunta@microsemi.com> Reviewed-by: David Carroll <David.Carroll@microsemi.com> Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
On I/O errors, the Windows driver doesn't set data_transfer_length
on error conditions other than SRB_STATUS_DATA_OVERRUN.
In these cases we need to set data_transfer_length to 0,
indicating there is no data transferred. On SRB_STATUS_DATA_OVERRUN,
data_transfer_length is set by the Windows driver to the actual data transferred.
Reported-by: Shiva Krishna <Shiva.Krishna@nimblestorage.com> Signed-off-by: Long Li <longli@microsoft.com> Reviewed-by: K. Y. Srinivasan <kys@microsoft.com> Signed-off-by: K. Y. Srinivasan <kys@microsoft.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
When sense message is present on error, we should pass along to the upper
layer to decide how to deal with the error.
This patch fixes connectivity issues with Fiber Channel devices.
Signed-off-by: Long Li <longli@microsoft.com> Reviewed-by: K. Y. Srinivasan <kys@microsoft.com> Signed-off-by: K. Y. Srinivasan <kys@microsoft.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Properly set SRB flags when hosting device supports tagged queuing.
This patch improves the performance on Fiber Channel disks.
Signed-off-by: Long Li <longli@microsoft.com> Reviewed-by: K. Y. Srinivasan <kys@microsoft.com> Signed-off-by: K. Y. Srinivasan <kys@microsoft.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The problem is that shmat() calls do_mmap_pgoff() with MAP_FIXED, and
the address rounded down to 0. For the regular mmap case, the
protection mentioned above is that the kernel gets to generate the
address -- arch_get_unmapped_area() will always check for MAP_FIXED and
return that address. So by the time we do security_mmap_addr(0) things
get funky for shmat().
The testcase itself shows that while a regular user crashes, root will
not have a problem attaching a nil-page. There are two possible fixes
to this. The first, and which this patch does, is to simply allow root
to crash as well -- this is also regular mmap behavior, ie when hacking
up the testcase and adding mmap(... |MAP_FIXED). While this approach
is the safer option, the second alternative is to ignore SHM_RND if the
rounded address is 0, thus only having MAP_SHARED flags. This makes the
behavior of shmat() identical to the mmap() case. The downside of this
is obviously user visible, but does make sense in that it maintains
semantics after the round-down wrt 0 address and mmap.
At the end of a window period, if the reclaimed pages is greater than
scanned, an unsigned underflow can result in a huge pressure value and
thus a critical event. Reclaimed pages is found to go higher than
scanned because of the addition of reclaimed slab pages to reclaimed in
shrink_node without a corresponding increment to scanned pages.
Minchan Kim mentioned that this can also happen in the case of a THP
page where the scanned is 1 and reclaimed could be 512.
Link: http://lkml.kernel.org/r/1486641577-11685-1-git-send-email-vinmenon@codeaurora.org Signed-off-by: Vinayak Menon <vinmenon@codeaurora.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Michal Hocko <mhocko@suse.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Rik van Riel <riel@redhat.com> Cc: Vladimir Davydov <vdavydov.dev@gmail.com> Cc: Anton Vorontsov <anton.vorontsov@linaro.org> Cc: Shiraz Hashim <shashim@codeaurora.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
gcc-7 detects that wlanhdr_to_ethhdr() in two drivers calls memcpy() with
a destination argument that an earlier function call may have set to NULL:
staging/rtl8188eu/core/rtw_recv.c: In function 'wlanhdr_to_ethhdr':
staging/rtl8188eu/core/rtw_recv.c:1318:2: warning: argument 1 null where non-null expected [-Wnonnull]
staging/rtl8712/rtl871x_recv.c: In function 'r8712_wlanhdr_to_ethhdr':
staging/rtl8712/rtl871x_recv.c:649:2: warning: argument 1 null where non-null expected [-Wnonnull]
I'm fixing this by adding a NULL pointer check and returning failure
from the function, which is hopefully already handled properly.
This seems to date back to when the drivers were originally added,
so backporting the fix to stable seems appropriate. There are other
related realtek drivers in the kernel, but none of them contain a
function with a similar name or produce this warning.
Fixes: 1cc18a22b96b ("staging: r8188eu: Add files for new driver - part 5") Fixes: 2865d42c78a9 ("staging: r8712u: Add the new driver to the mainline kernel") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The sequencer FIFO management has a bug that may lead to a corruption
(shortage) of the cell linked list. When a sequencer client faces an
error at the event delivery, it tries to put back the dequeued cell.
When the first queue was put back, this forgot the tail pointer
tracking, and the link will be screwed up.
Although there is no memory corruption, the sequencer client may stall
forever at exit while flushing the pending FIFO cells in
snd_seq_pool_done(), as spotted by syzkaller.
This patch addresses the missing tail pointer tracking at
snd_seq_fifo_cell_putback(). Also the patch makes sure to clear the
cell->enxt pointer at snd_seq_fifo_event_in() for avoiding a similar
mess-up of the FIFO linked list.
When a user sets a too small ticks with a fine-grained timer like
hrtimer, the kernel tries to fire up the timer irq too frequently.
This may lead to the condensed locks, eventually the kernel spinlock
lockup with warnings.
For avoiding such a situation, we define a lower limit of the
resolution, namely 1ms. When the user passes a too small tick value
that results in less than that, the kernel returns -EINVAL now.
Like for Sunrise Point, the total stream number of Lewisburg's
input and output stream exceeds 15 (GCAP is 0x9701), which will
cause some streams do not work because of the overflow on
SDxCTL.STRM field if using the legacy stream tag allocation method.
is_jump_ins() checks for plain jump ("j") instructions since commit e7438c4b893e ("MIPS: Fix sibling call handling in get_frame_info") but
that commit didn't make the same change to the microMIPS code, leaving
it inconsistent with the MIPS32/MIPS64 code. Handle the microMIPS
encoding of the jump instruction too such that it behaves consistently.
Signed-off-by: Paul Burton <paul.burton@imgtec.com> Fixes: e7438c4b893e ("MIPS: Fix sibling call handling in get_frame_info") Cc: Tony Wu <tung7970@gmail.com> Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/14533/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
get_frame_info() calculates the offset of the return address within a
stack frame simply by dividing a the bottom 16 bits of the instruction,
treated as a signed integer, by the size of a long. Whilst this works
for MIPS32 & MIPS64 ISAs where the sw or sd instructions are used, it's
incorrect for microMIPS where encodings differ. The result is that we
typically completely fail to unwind the stack on microMIPS.
Fix this by adjusting is_ra_save_ins() to calculate the return address
offset, and take into account the various different encodings there in
the same place as we consider whether an instruction is storing the
ra/$31 register.
With this we are now able to unwind the stack for kernels targetting the
microMIPS ISA, for example we can produce:
is_jump_ins() checks 16b instruction fields without verifying that the
instruction is indeed 16b, as is done by is_ra_save_ins() &
is_sp_move_ins(). Add the appropriate check.
get_frame_info() is meant to iterate over up to the first 128
instructions within a function, but for microMIPS kernels it will not
reach that many instructions unless the function is 512 bytes long since
we calculate the maximum number of instructions to check by dividing the
function length by the 4 byte size of a union mips_instruction. In
microMIPS kernels this won't do since instructions are variable length.
Fix this by instead checking whether the pointer to the current
instruction has reached the end of the function, and use max_insns as a
simple constant to check the number of iterations against.
During stack unwinding we call a number of functions to determine what
type of instruction we're looking at. The union mips_instruction pointer
provided to them may be pointing at a 2 byte, but not 4 byte, aligned
address & we thus cannot directly access the 4 byte wide members of the
union mips_instruction. To avoid this is_ra_save_ins() copies the
required half-words of the microMIPS instruction to a correctly aligned
union mips_instruction on the stack, which it can then access safely.
The is_jump_ins() & is_sp_move_ins() functions do not correctly perform
this temporary copy, and instead attempt to directly dereference 4 byte
fields which may be misaligned and lead to an address exception.
Fix this by copying the instruction halfwords to a temporary union
mips_instruction in get_frame_info() such that we can provide a 4 byte
aligned union mips_instruction to the is_*_ins() functions and they do
not need to deal with misalignment themselves.
get_frame_info() can be called in microMIPS kernels with the ISA bit
already clear. For example this happens when unwind_stack_by_address()
is called because we begin with a PC that has the ISA bit set & subtract
the (odd) offset from the preceding symbol (which does not have the ISA
bit set). Since get_frame_info() unconditionally subtracts 1 from the PC
in microMIPS kernels it incorrectly misaligns the address it then
attempts to access code at, leading to an address error exception.
Fix this by using msk_isa16_mode() to clear the ISA bit, which allows
get_frame_info() to function regardless of whether it is provided with a
PC that has the ISA bit set or not.
If copy_from_user is called with a large buffer (>= 128 bytes) and the
userspace buffer refers partially to unreadable memory, then it is
possible for Octeon's copy_from_user to report the wrong number of bytes
have been copied. In the case where the buffer size is an exact multiple
of 128 and the fault occurs in the last 64 bytes, copy_from_user will
report that all the bytes were copied successfully but leave some
garbage in the destination buffer.
The bug is in the main __copy_user_common loop in octeon-memcpy.S where
in the middle of the loop, src and dst are incremented by 128 bytes. The
l_exc_copy fault handler is used after this but that assumes that
"src < THREAD_BUADDR($28)". This is not the case if src has already been
incremented.
Fix by adding an extra fault handler which rewinds the src and dst
pointers 128 bytes before falling though to l_exc_copy.
Thanks to the pwritev test from the strace test suite for originally
highlighting this bug!
Fixes: 5b3b16880f40 ("MIPS: Add Cavium OCTEON processor support ...") Signed-off-by: James Cowgill <James.Cowgill@imgtec.com> Acked-by: David Daney <david.daney@cavium.com> Reviewed-by: James Hogan <james.hogan@imgtec.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/14978/ Signed-off-by: James Hogan <james.hogan@imgtec.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
For certain arguments such as saddr = 0xc0a8fd60, daddr = 0xc0a8fda1,
len = 80, proto = 17, sum = 0x7eae049d there will be a carry when
folding the intermediate 64 bit checksum to 32 bit but the code doesn't
add the carry back to the one's complement sum, thus an incorrect result
will be generated.
Reported-by: Mark Zhang <bomb.zhang@gmail.com> Signed-off-by: Ralf Baechle <ralf@linux-mips.org> Reviewed-by: James Hogan <james.hogan@imgtec.com> Signed-off-by: James Hogan <james.hogan@imgtec.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Instead of the one when another syscall takes place while another is being
processed (in another CPU, but we show it serialized, so need to "interrupt"
the other), and also when finally showing the sys_enter + sys_exit + duration,
where we were showing the sample->time for the sys_exit, duh.
[js] no trace__printf_interrupted_entry in 3.12 yet
Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/n/tip-ecbzgmu2ni6glc6zkw8p1zmx@git.kernel.org Fixes: 752fde44fd1c ("perf trace: Support interrupted syscalls") Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
META_COLLECTOR int_vlan_tag() assumes that if the accel tag (vlan_tci)
is zero, then no vlan accel tag is present.
This is incorrect for zero VID vlan accel packets, making the following
match fail:
tc filter add ... basic match 'meta(vlan mask 0xfff eq 0)' ...
Apparently 'int_vlan_tag' was implemented prior VLAN_TAG_PRESENT was
introduced in 05423b2 "vlan: allow null VLAN ID to be used"
(and at time introduced, the 'vlan_tx_tag_get' call in em_meta was not
adapted).
Fix, testing skb_vlan_tag_present instead of testing skb_vlan_tag_get's
value.
Fixes: 05423b2413 ("vlan: allow null VLAN ID to be used") Fixes: 1a31f2042e ("netsched: Allow meta match on vlan tag on receive") Signed-off-by: Shmulik Ladkani <shmulik.ladkani@gmail.com> Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
commit 2dcab5984841 ("sctp: avoid BUG_ON on sctp_wait_for_sndbuf")
attempted to avoid a BUG_ON call when the association being used for a
sendmsg() is blocked waiting for more sndbuf and another thread did a
peeloff operation on such asoc, moving it to another socket.
As Ben Hutchings noticed, then in such case it would return without
locking back the socket and would cause two unlocks in a row.
Further analysis also revealed that it could allow a double free if the
application managed to peeloff the asoc that is created during the
sendmsg call, because then sctp_sendmsg() would try to free the asoc
that was created only for that call.
This patch takes another approach. It will deny the peeloff operation
if there is a thread sleeping on the asoc, so this situation doesn't
exist anymore. This avoids the issues described above and also honors
the syscalls that are already being handled (it can be multiple sendmsg
calls).
Joint work with Xin Long.
Fixes: 2dcab5984841 ("sctp: avoid BUG_ON on sctp_wait_for_sndbuf") Cc: Alexander Popov <alex.popov@linux.com> Cc: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: Xin Long <lucien.xin@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Recently I receive a bug report that on Linux v3.0 based kerenl, hot add
disk to a md linear device causes kernel crash at linear_congested(). From
the crash image analysis, I find in linear_congested(), mddev->raid_disks
contains value N, but conf->disks[] only has N-1 pointers available. Then
a NULL pointer deference crashes the kernel.
There is a race between linear_add() and linear_congested(), RCU stuffs
used in these two functions cannot avoid the race. Since Linuv v4.0
RCU code is replaced by introducing mddev_suspend(). After checking the
upstream code, it seems linear_congested() is not called in
generic_make_request() code patch, so mddev_suspend() cannot provent it
from being called. The possible race still exists.
Here I explain how the race still exists in current code. For a machine
has many CPUs, on one CPU, linear_add() is called to add a hard disk to a
md linear device; at the same time on other CPU, linear_congested() is
called to detect whether this md linear device is congested before issuing
an I/O request onto it.
Now I use a possible code execution time sequence to demo how the possible
race happens,
In linear_add() mddev->raid_disks is increased in time seq 2, and on
another CPU in linear_congested() the for-loop iterates conf->disks[i] by
the increased mddev->raid_disks in time seq 3,4. But conf with one more
element (which is a pointer to struct dev_info type) to conf->disks[] is
not updated yet, accessing its structure member in time seq 4 will cause a
NULL pointer deference fault.
To fix this race, there are 2 parts of modification in the patch,
1) Add 'int raid_disks' in struct linear_conf, as a copy of
mddev->raid_disks. It is initialized in linear_conf(), always being
consistent with pointers number of 'struct dev_info disks[]'. When
iterating conf->disks[] in linear_congested(), use conf->raid_disks to
replace mddev->raid_disks in the for-loop, then NULL pointer deference
will not happen again.
2) RCU stuffs are back again, and use kfree_rcu() in linear_add() to
free oldconf memory. Because oldconf may be referenced as mddev->private
in linear_congested(), kfree_rcu() makes sure that its memory will not
be released until no one uses it any more.
Also some code comments are added in this patch, to make this modification
to be easier understandable.
This patch can be applied for kernels since v4.0 after commit: 3be260cc18f8 ("md/linear: remove rcu protections in favour of
suspend/resume"). But this bug is reported on Linux v3.0 based kernel, for
people who maintain kernels before Linux v4.0, they need to do some back
back port to this patch.
Changelog:
- V3: add 'int raid_disks' in struct linear_conf, and use kfree_rcu() to
replace rcu_call() in linear_add().
- v2: add RCU stuffs by suggestion from Shaohua and Neil.
- v1: initial effort.
Signed-off-by: Coly Li <colyli@suse.de> Cc: Shaohua Li <shli@fb.com> Cc: Neil Brown <neilb@suse.com> Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Fix errors during open not being returned to userspace. Specifically,
failed control-line manipulations or control or read urb submissions
would not be detected.
If acm_submit_read_urbs() fails in acm_port_activate(), error handling
code calls usb_autopm_put_interface() while it is already called
before acm_submit_read_urbs(). The patch reorganizes error handling code
to avoid double decrement of USB interface's PM-usage counter.
Found by Linux Driver Verification project (linuxtesting.org).
This is the s390 variant of Alexei's JIT bug fix.
(patch description below stolen from Alexei's patch)
bpf_alloc_binary() adds 128 bytes of room to JITed program image
and rounds it up to the nearest page size. If image size is close
to page size (like 4000), it is rounded to two pages:
round_up(4000 + 4 + 128) == 8192
then 'hole' is computed as 8192 - (4000 + 4) = 4188
If prandom_u32() % hole selects a number >= PAGE_SIZE - sizeof(*header)
then kernel will crash during bpf_jit_free():
since bpf_jit_free() does:
unsigned long addr = (unsigned long)fp->bpf_func & PAGE_MASK;
struct bpf_binary_header *header = (void *)addr;
to compute start address of 'bpf_binary_header'
and header->pages will pass junk to:
set_memory_rw(addr, header->pages);
Fix it by making sure that &header->image[prandom_u32() % hole] and &header
are in the same page.
Fixes: aa2d2c73c21f2 ("s390/bpf,jit: address randomize and write protect jit code") Reported-by: Alexei Starovoitov <ast@plumgrid.com> Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
bpf_alloc_binary() adds 128 bytes of room to JITed program image
and rounds it up to the nearest page size. If image size is close
to page size (like 4000), it is rounded to two pages:
round_up(4000 + 4 + 128) == 8192
then 'hole' is computed as 8192 - (4000 + 4) = 4188
If prandom_u32() % hole selects a number >= PAGE_SIZE - sizeof(*header)
then kernel will crash during bpf_jit_free():
since bpf_jit_free() does:
unsigned long addr = (unsigned long)fp->bpf_func & PAGE_MASK;
struct bpf_binary_header *header = (void *)addr;
to compute start address of 'bpf_binary_header'
and header->pages will pass junk to:
set_memory_rw(addr, header->pages);
Fix it by making sure that &header->image[prandom_u32() % hole] and &header
are in the same page
Fixes: 314beb9bcabfd ("x86: bpf_jit_comp: secure bpf jit against spraying attacks") Signed-off-by: Alexei Starovoitov <ast@plumgrid.com> Acked-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
This fixes a seg fault on 'ethtool -A' entry if the
interface is down. Obviously we need to have the
phy device initialized / "connected" (see of_phy_connect())
to be able to advertise pause frame capabilities.
Fixes: 23402bddf9e56eecb27bbd1e5467b3b79b3dbe58 Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
MBus windows are used on Marvell platforms to map certain peripherals
in the physical address space. In the PCIe context, MBus windows are
needed to map PCIe I/O and memory regions in the physical address.
However, those MBus windows can only have power of two sizes, while
PCIe BAR do not necessarily guarantee this. For this reason, the
current pci-mvebu breaks on platforms where PCIe devices have BARs
that don't sum up to a power of two size at the emulated bridge level.
This commit fixes this by allowing the pci-mvebu driver to create
multiple contiguous MBus windows (each having a power of two size) to
cover a given PCIe BAR.
To achieve this, two functions are added: mvebu_pcie_add_windows() and
mvebu_pcie_del_windows() to respectively add and remove all the MBus
windows that are needed to map the provided PCIe region base and
size. The emulated PCI bridge code now calls those functions, instead
of directly calling the mvebu-mbus driver functions.
Fixes: 45361a4fe446 ('pci: PCIe driver for Marvell Armada 370/XP systems') Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Link: https://lkml.kernel.org/r/1397823593-1932-8-git-send-email-thomas.petazzoni@free-electrons.com Tested-by: Neil Greatorex <neil@fatboyfat.co.uk> Acked-by: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
projects / thirdparty / kernel / stable.git / log
