wd719x_template is missing the .module field, causing module refcount
not to work, allowing to rmmod the driver while in use (mounted filesystem),
causing an oops.
Set .module to THIS_MODULE to fix the problem.
Signed-off-by: Ondrej Zary <linux@rainbow-software.org> Signed-off-by: James Bottomley <JBottomley@Parallels.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Each inode of nilfs2 stores a root node of a b-tree, and it turned out to
have a memory overrun issue:
Each b-tree node of nilfs2 stores a set of key-value pairs and the number
of them (in "bn_nchildren" member of nilfs_btree_node struct), as well as
a few other "bn_*" members.
Since the value of "bn_nchildren" is used for operations on the key-values
within the b-tree node, it can cause memory access overrun if a large
number is incorrectly set to "bn_nchildren".
For instance, nilfs_btree_node_lookup() function determines the range of
binary search with it, and too large "bn_nchildren" leads
nilfs_btree_node_get_key() in that function to overrun.
As for intermediate b-tree nodes, this is prevented by a sanity check
performed when each node is read from a drive, however, no sanity check
has been done for root nodes stored in inodes.
This patch fixes the issue by adding missing sanity check against b-tree
root nodes so that it's called when on-memory inodes are read from ifile,
inode metadata file.
When the last on-demand paging MR is released the notifier count is
left non-zero so that concurrent page faults will have to abort. If a
new MR is then registered, the counter is reset. However, the decision
is made to put the new MR in the list waiting for the notifier count
to reach zero, before the counter is reset. An invalidation or another
MR registration can release the MR to handle page faults, but without
such an event the MR can wait forever.
The patch fixes this issue by adding a check whether the MR is the
first on-demand paging MR when deciding whether it is ready to handle
page faults. If it is the first MR, we know that there are no mmu
notifiers running in parallel to the registration.
Fixes: 882214e2b128 ("IB/core: Implement support for MMU notifiers regarding on demand paging regions") Signed-off-by: Haggai Eran <haggaie@mellanox.com> Signed-off-by: Shachar Raindel <raindel@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The deadlock occurs in __uverbs_modify_qp: we take a lock (idr_read_qp)
and in case of failure in ib_resolve_eth_l2_attrs we don't release
it (put_qp_read). Fix that.
Fixes: ed4c54e5b4ba ("IB/core: Resolve Ethernet L2 addresses when modifying QP") Signed-off-by: Moshe Lazer <moshel@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The MLX4_PROT_IB_IPV4 protocol should only be used with RoCEv2 and such.
Removing this wrong usage allows to run multicast applications over RoCE.
Fixes: d487ee77740c ("IB/mlx4: Use IBoE (RoCE) IP based GIDs in the port GID table") Reported-by: Carol Soto <clsoto@linux.vnet.ibm.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In case handle_eth_ud_smac_index fails, we need to free the allocated resources.
Fixes: 2f5bb473681b ("mlx4: Add ref counting to port MAC table for RoCE") Signed-off-by: Majd Dibbiny <majd@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When teardown process starts during live IO, we need to keep the
memory regions pool (frmr/fmr) until all in-flight tasks are properly
released, since each task may return a memory region to the pool. In
order to do this, we pass a destroy flag to iser_free_ib_conn_res to
indicate we can destroy the device and the memory regions
pool. iser_conn_release will pass it as true and also DEVICE_REMOVAL
event (we need to let the device to properly remove).
Also, Since we conditionally call iser_free_rx_descriptors,
remove the extra check on iser_conn->rx_descs.
Fixes: 5426b1711fd0 ("IB/iser: Collapse cleanup and disconnect handlers") Reported-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Sagi Grimberg <sagig@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This changeset removes all the code that allows the driver to write to
the EEPROM and update the recorded error counters and power on hours.
These two stats are unused and writing them exposes a timing risk
which could leave the EEPROM in a bad state preventing further normal
operation of the HCA.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
With task_blocks_on_rt_mutex() returning early -EDEADLK we never
add the waiter to the waitqueue. Later, we try to remove it via
remove_waiter() and go boom in rt_mutex_top_waiter() because
rb_entry() gives a NULL pointer.
( Tested on v3.18-RT where rtmutex is used for regular mutex and I
tried to get one twice in a row. )
Not sure when this started but I guess 397335f004f4 ("rtmutex: Fix
deadlock detector for real") or commit 3d5c9340d194 ("rtmutex:
Handle deadlock detection smarter").
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Acked-by: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1424187823-19600-1-git-send-email-bigeasy@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This is essentially a partial revert of the commit [b1920c21102a:
'ALSA: hda - Enable runtime PM on Panther Point']. There was a bug
report showing the HD-audio bus hang during runtime PM on HP Spectre
XT.
A part of these drivers, especially BeBoB driver, are programmed to wait
some events. Thus the drivers should not destroy any data in .remove()
context.
This commit moves some destructors from 'struct fw_driver.remove()' to
'struct snd_card.private_free()' to shutdown safely.
Currently stream destructor in each driver has a problem to be called in
a context in which sound card object is released, because the destructors
call amdtp_stream_pcm_abort() and touch PCM runtime data.
The PCM runtime data is destroyed in application's context with
snd_pcm_close(), on the other hand PCM substream data is destroyed after
sound card object is released, in most case after all of ALSA character
devices are released. When PCM runtime is destroyed and PCM substream is
remained, amdtp_stream_pcm_abort() touches PCM runtime data and causes
Null-pointer-dereference.
This commit changes stream destructors and allows each driver to call
it after releasing runtime.
AMDTP helper functions increment/decrement reference counter for an
instance of FireWire unit, while it's complicated for each driver to
process error state.
In previous commit, each driver has the role of reference counting. This
commit removes this role from the helper function.
Fireworks and Dice drivers try to touch instances of FireWire unit after
sound card object is released, while references to the unit is decremented
in .remove(). When unplugging during streaming, sound card object is
released after .remove(), thus Fireworks and Dice drivers causes GPF or
Null-pointer-dereferencing to application processes because an instance of
FireWire unit was already released.
This commit adds reference-counting for FireWire unit in drivers to allow
them to touch an instance of FireWire unit after .remove(). In most case,
any operations after .remove() may be failed safely.
When a PCM draining is performed to an empty stream that has been
already in PREPARED state, the current code just ignores and leaves as
it is, although the drain is supposed to set all such streams to SETUP
state. This patch covers that overlooked case.
The conditional RX-FIFO read seems to cause spurious interrupts and we
see just:
|serial8250: too much work for irq29
The previous behaviour was "default" for decades and Marvell's 88f6282 SoC
might not be the only that relies on it. Therefore the Omap fix is
reverted for now.
Fixes: 0aa525d11859 ("tty: serial: 8250_core: read only RX if there is
something in the FIFO") Reported-By: Nicolas Schichan <nschichan@freebox.fr> Debuged-By: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
But it still misses one point. As John Paul correctly points out, we
do not care about setting date. If somebody ever changes wall
time backwards (by mistake for example), tty timestamps are never
updated until the original wall time passes.
So check the absolute difference of times and if it large than "8
seconds or so", always update the time. That means we will update
immediatelly when changing time. Ergo, CAP_SYS_TIME can foul the
check, but it was always that way.
Thanks John for serving me this so nicely debugged.
Signed-off-by: Jiri Slaby <jslaby@suse.cz> Reported-by: John Paul Perry <john_paul.perry@alcatel-lucent.com> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
/proc/<pid>/maps currently don't annotate stack vma with "[stack]"
This is because KSTK_ESP ie expected to return usermode SP of tsk while
currently it returns the kernel mode SP of a sleeping tsk.
While the fix is trivial, we also need to adjust the ARC kernel stack
unwinder to not use KSTK_SP and friends any more.
As it is, we have debugfs_remove() racing with symlink traversals.
Supply ->evict_inode() and do freeing there - inode will remain
pinned until we are done with the symlink body.
And rip the idiocy with checking if dentry is positive right after
we'd verified debugfs_positive(), which is a stronger check...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add missing error handling when registering the tty device at port
probe. This avoids trying to remove an uninitialised character device
when the port device is removed.
Fix return value in probe error path, which could end up returning
success (0) on errors. This could in turn lead to use-after-free or
double free (e.g. in port_remove) when the port device is removed.
Fixes: c706ebdfc895 ("USB: usb-serial: call port_probe and port_remove
at the right times") Signed-off-by: Johan Hovold <johan@kernel.org> Acked-by: Greg Kroah-Hartman <greg@kroah.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fix overflow bug in tty_wait_until_sent on 64-bit machines, where an
infinite timeout (0) would be passed to the underlying tty-driver's
wait_until_sent-operation as a negative timeout (-1), causing it to
return immediately.
This manifests itself for example as tcdrain() returning immediately,
drivers not honouring the drain flags when setting terminal attributes,
or even dropped data on close as a requested infinite closing-wait
timeout would be ignored.
The first symptom was reported by Asier LLANO who noted that tcdrain()
returned prematurely when using the ftdi_sio usb-serial driver.
Fix this by passing 0 rather than MAX_SCHEDULE_TIMEOUT (LONG_MAX) to the
underlying tty driver.
Note that the serial-core wait_until_sent-implementation is not affected
by this bug due to a lucky chance (comparison to an unsigned maximum
timeout), and neither is the cyclades one that had an explicit check for
negative timeouts, but all other tty drivers appear to be affected.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Reported-by: ZIV-Asier Llano Palacios <asier.llano@cgglobal.com> Signed-off-by: Johan Hovold <johan@kernel.org> Reviewed-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The call to cfg80211_ch_switch_notify() should be at the end of the
ieee80211_chswitch_post_beacon() function, because it should only be
sent if everything succeeded.
Fixes: d04b5ac9e70b ("cfg80211/mac80211: allow any interface to send channel switch notifications") Signed-off-by: Luciano Coelho <luciano.coelho@intel.com> Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The current minstrel_ht rate control behavior is somewhat optimistic in
trying to find optimum TX rate. While this is usually fine for normal
Data frames, there are cases where a more conservative set of retry
parameters would be beneficial to make the connection more robust.
EAPOL frames are critical to the authentication and especially the
EAPOL-Key message 4/4 (the last message in the 4-way handshake) is
important to get through to the AP. If that message is lost, the only
recovery mechanism in many cases is to reassociate with the AP and start
from scratch. This can often be avoided by trying to send the frame with
more conservative rate and/or with more link layer retries.
In most cases, minstrel_ht is currently using the initial EAPOL-Key
frames for probing higher rates and this results in only five link layer
transmission attempts (one at high(ish) MCS and four at MCS0). While
this works with most APs, it looks like there are some deployed APs that
may have issues with the EAPOL frames using HT MCS immediately after
association. Similarly, there may be issues in cases where the signal
strength or radio environment is not good enough to be able to get
frames through even at couple of MCS 0 tries.
The best approach for this would likely to be to reduce the TX rate for
the last rate (3rd rate parameter in the set) to a low basic rate (say,
6 Mbps on 5 GHz and 2 or 5.5 Mbps on 2.4 GHz), but doing that cleanly
requires some more effort. For now, we can start with a simple one-liner
that forces the minimum rate to be used for EAPOL frames similarly how
the TX rate is selected for the IEEE 802.11 Management frames. This does
result in a small extra latency added to the cases where the AP would be
able to receive the higher rate, but taken into account how small number
of EAPOL frames are used, this is likely to be insignificant. A future
optimization in the minstrel_ht design can also allow this patch to be
reverted to get back to the more optimized initial TX rate.
It should also be noted that many drivers that do not use minstrel as
the rate control algorithm are already doing similar workarounds by
forcing the lowest TX rate to be used for EAPOL frames.
The xhci in Intel Sunrisepoint and Cherryview platforms need a driver
workaround for a Stuck PME that might either block PME events in suspend,
or create spurious PME events preventing runtime suspend.
Workaround is to clear a internal PME flag, BIT(28) in a vendor specific
PMCTRL register at offset 0x80a4, in both suspend resume callbacks
Without this, xhci connected usb devices might never be able to wake up the
system from suspend, or prevent device from going to suspend (xhci d3)
When a control transfer has a short data stage, the xHCI controller generates
two transfer events: a COMP_SHORT_TX event that specifies the untransferred
amount, and a COMP_SUCCESS event. But when the data stage is not short, only the
COMP_SUCCESS event occurs. Therefore, xhci-hcd must set urb->actual_length to
urb->transfer_buffer_length while processing the COMP_SUCCESS event, unless
urb->actual_length was set already by a previous COMP_SHORT_TX event.
The driver checks this by seeing whether urb->actual_length == 0, but this alone
is the wrong test, as it is entirely possible for a short transfer to have an
urb->actual_length = 0.
This patch changes the xhci driver to rely on a new td->urb_length_set flag,
which is set to true when a COMP_SHORT_TX event is received and the URB length
updated at that stage.
This fixes a bug which affected the HSO plugin, which relies on URBs with
urb->actual_length == 0 to halt re-submitting the RX URB in the control
endpoint.
Include the high order bit fields for Max scratchpad buffers when
calculating how many scratchpad buffers are needed.
I'm suprised this hasn't caused more issues, we never allocated more than
32 buffers even if xhci needed more. Either we got lucky and xhci never
really used past that area, or then we got enough zeroed dma memory anyway.
Should be backported as far back as possible
Reported-by: Tim Chen <tim.c.chen@linux.intel.com> Tested-by: Tim Chen <tim.c.chen@linux.intel.com> Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The commit 973747928514 ("usb: host: xhci-plat: add support for the Armada
375/38x XHCI controllers") extended the xhci-plat driver to support the Armada
375/38x SoCs, mostly by adding a quirk configuring the MBUS window.
However, that quirk was run before the clock the controllers needs has been
enabled. This usually worked because the clock was first enabled by the
bootloader, and left as such until the driver is probe, where it tries to
access the MBUS configuration registers before enabling the clock.
Things get messy when EPROBE_DEFER is involved during the probe, since as part
of its error path, the driver will rightfully disable the clock. When the
driver will be reprobed, it will retry to access the MBUS registers, but this
time with the clock disabled, which hangs forever.
Fix this by running the quirks after the clock has been enabled by the driver.
The "Extended Compat ID OS Feature Descriptor Specification" does not
require the (sub)compatible ids to be NUL-terminated, because they
are placed in a fixed-size buffer and only unused parts of it should
contain NULs. If the buffer is fully utilized, there is no place for NULs.
Consequently, the code which uses desc->ext_compat_id never expects the
data contained to be NUL terminated.
If the compatible id is stored after sub-compatible id, and the compatible
id is full length (8 bytes), the (useless) NUL terminator overwrites the
first byte of the sub-compatible id.
If the sub-compatible id is full length (8 bytes), the (useless) NUL
terminator ends up out of the buffer. The situation can happen in the RNDIS
function, where the buffer is a part of struct f_rndis_opts. The next
member of struct f_rndis_opts is a mutex, so its first byte gets
overwritten. The said byte is a part of a mutex'es member which contains
the information on whether the muext is locked or not. This can lead to a
deadlock, because, in a configfs-composed gadget when a function is linked
into a configuration with config_usb_cfg_link(), usb_get_function()
is called, which then calls rndis_alloc(), which tries locking the same
mutex and (wrongly) finds it already locked.
This patch eliminates NUL terminating of the (sub)compatible id.
Fixes: da4243145fb1: "usb: gadget: configfs: OS Extended Compatibility descriptors support" Reported-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Andrzej Pietrasiewicz <andrzej.p@samsung.com> Signed-off-by: Felipe Balbi <balbi@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In the wrapper the IRQ disable should be done by writing 1's to the
IRQ*_CLR register. Existing code is broken because it instead writes
zeros to IRQ*_SET register.
Fix this by adding functions dwc3_omap_write_irqmisc_clr() and
dwc3_omap_write_irq0_clr() which do the right thing.
Fixes: 72246da40f37 ("usb: Introduce DesignWare USB3 DRD Driver") Signed-off-by: George Cherian <george.cherian@ti.com> Signed-off-by: Felipe Balbi <balbi@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch integrates Cyber Cortex AV boards with the existing
ftdi_jtag_quirk in order to use serial port 0 with JTAG which is
required by the manufacturers' software.
Steps: 2
[ftdi_sio_ids.h]
1. Defined the device PID
[ftdi_sio.c]
2. Added a macro declaration to the ids array, in order to enable the
jtag quirk for the device.
Signed-off-by: Max Mansfield <max.m.mansfield@gmail.com> Signed-off-by: Johan Hovold <johan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
These product identifiers (PID) all deal with marine NMEA format data
used on motor boats and yachts. We supply the programmed devices to
Chetco, for use inside their equipment. The PIDs are a direct copy of
our Windows device drivers (FTDI drivers with altered PIDs).
Signed-off-by: Mark Glover <mark@actisense.com>
[johan: edit commit message slightly ] Signed-off-by: Johan Hovold <johan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When a signal is delivered, the information in the siginfo structure
is copied to userspace. Good security practice dicatates that the
unused fields in this structure should be initialized to 0 so that
random kernel stack data isn't exposed to the user. This patch adds
such an initialization to the two places where usbfs raises signals.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Reported-by: Dave Mielke <dave@mielke.cc> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
These device ID's are not associated with the cp210x module currently,
but should be. This patch allows the devices to operate upon connecting
them to the usb bus as intended.
Signed-off-by: Michiel van de Garde <mgparser@gmail.com> Signed-off-by: Johan Hovold <johan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A bulk-out size smaller than the end-point size is indeed valid. The
offending commit broke the usb-debug driver for EHCI debug devices,
which use 8-byte buffers.
Fixes: 5083fd7bdfe6 ("USB: serial: make bulk_out_size a lower limit") Reported-by: "Li, Elvin" <elvin.li@intel.com> Signed-off-by: Johan Hovold <johan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Like the JMicron JMS567 enclosures with the JMS539 choke on report-opcodes,
so avoid it.
Tested-and-reported-by: Tom Arild Naess <tanaess@gmail.com> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently the guest exit trace event saves the VCPU pointer to the
structure, and the guest PC is retrieved by dereferencing it when the
event is printed rather than directly from the trace record. This isn't
safe as the printing may occur long afterwards, after the PC has changed
and potentially after the VCPU has been freed. Usually this results in
the same (wrong) PC being printed for multiple trace events. It also
isn't portable as userland has no way to access the VCPU data structure
when interpreting the trace record itself.
Lets save the actual PC in the structure so that the correct value is
accessible later.
Fixes: 669e846e6c4e ("KVM/MIPS32: MIPS arch specific APIs for KVM") Signed-off-by: James Hogan <james.hogan@imgtec.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Gleb Natapov <gleb@kernel.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: linux-mips@linux-mips.org Cc: kvm@vger.kernel.org Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This has been broken for a long time: it broke first in 2.6.35, then was
almost fixed in 2.6.36 but this one-liner slipped through the cracks.
The bug shows up as an infinite loop in Windows 7 (and newer) boot on
32-bit hosts without EPT.
Windows uses CMPXCHG8B to write to page tables, which causes a
page fault if running without EPT; the emulator is then called from
kvm_mmu_page_fault. The loop then happens if the higher 4 bytes are
not 0; the common case for this is that the NX bit (bit 63) is 1.
Fixes: 6550e1f165f384f3a46b60a1be9aba4bc3c2adad Fixes: 16518d5ada690643453eb0aef3cc7841d3623c2d Reported-by: Erik Rull <erik.rull@rdsoftware.de> Tested-by: Erik Rull <erik.rull@rdsoftware.de> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When using the fast file fsync code path we can miss the fact that new
writes happened since the last file fsync and therefore return without
waiting for the IO to finish and write the new extents to the fsync log.
Here's an example scenario where the fsync will miss the fact that new
file data exists that wasn't yet durably persisted:
1. fs_info->last_trans_committed == N - 1 and current transaction is
transaction N (fs_info->generation == N);
2. do a buffered write;
3. fsync our inode, this clears our inode's full sync flag, starts
an ordered extent and waits for it to complete - when it completes
at btrfs_finish_ordered_io(), the inode's last_trans is set to the
value N (via btrfs_update_inode_fallback -> btrfs_update_inode ->
btrfs_set_inode_last_trans);
4. transaction N is committed, so fs_info->last_trans_committed is now
set to the value N and fs_info->generation remains with the value N;
5. do another buffered write, when this happens btrfs_file_write_iter
sets our inode's last_trans to the value N + 1 (that is
fs_info->generation + 1 == N + 1);
6. transaction N + 1 is started and fs_info->generation now has the
value N + 1;
7. transaction N + 1 is committed, so fs_info->last_trans_committed
is set to the value N + 1;
8. fsync our inode - because it doesn't have the full sync flag set,
we only start the ordered extent, we don't wait for it to complete
(only in a later phase) therefore its last_trans field has the
value N + 1 set previously by btrfs_file_write_iter(), and so we
have:
Which made us not log the last buffered write and exit the fsync
handler immediately, returning success (0) to user space and resulting
in data loss after a crash.
This can actually be triggered deterministically and the following excerpt
from a testcase I made for xfstests triggers the issue. It moves a dummy
file across directories and then fsyncs the old parent directory - this
is just to trigger a transaction commit, so moving files around isn't
directly related to the issue but it was chosen because running 'sync' for
example does more than just committing the current transaction, as it
flushes/waits for all file data to be persisted. The issue can also happen
at random periods, since the transaction kthread periodicaly commits the
current transaction (about every 30 seconds by default).
The body of the test is:
# Create our main test file 'foo', the one we check for data loss.
# By doing an fsync against our file, it makes btrfs clear the 'needs_full_sync'
# bit from its flags (btrfs inode specific flags).
$XFS_IO_PROG -f -c "pwrite -S 0xaa 0 8K" \
-c "fsync" $SCRATCH_MNT/foo | _filter_xfs_io
# Now create one other file and 2 directories. We will move this second file
# from one directory to the other later because it forces btrfs to commit its
# currently open transaction if we fsync the old parent directory. This is
# necessary to trigger the data loss bug that affected btrfs.
mkdir $SCRATCH_MNT/testdir_1
touch $SCRATCH_MNT/testdir_1/bar
mkdir $SCRATCH_MNT/testdir_2
# Make sure everything is durably persisted.
sync
# Write more 8Kb of data to our file.
$XFS_IO_PROG -c "pwrite -S 0xbb 8K 8K" $SCRATCH_MNT/foo | _filter_xfs_io
# Move our 'bar' file into a new directory.
mv $SCRATCH_MNT/testdir_1/bar $SCRATCH_MNT/testdir_2/bar
# Fsync our first directory. Because it had a file moved into some other
# directory, this made btrfs commit the currently open transaction. This is
# a condition necessary to trigger the data loss bug.
$XFS_IO_PROG -c "fsync" $SCRATCH_MNT/testdir_1
# Now fsync our main test file. If the fsync succeeds, we expect the 8Kb of
# data we wrote previously to be persisted and available if a crash happens.
# This did not happen with btrfs, because of the transaction commit that
# happened when we fsynced the parent directory.
$XFS_IO_PROG -c "fsync" $SCRATCH_MNT/foo
# Simulate a crash/power loss.
_load_flakey_table $FLAKEY_DROP_WRITES
_unmount_flakey
# Now check that all data we wrote before are available.
echo "File content after log replay:"
od -t x1 $SCRATCH_MNT/foo
status=0
exit
The expected golden output for the test, which is what we get with this
fix applied (or when running against ext3/4 and xfs), is:
wrote 8192/8192 bytes at offset 0
XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
wrote 8192/8192 bytes at offset 8192
XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
File content after log replay: 0000000 aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa
* 0020000 bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb bb
* 0040000
Without this fix applied, the output shows the test file does not have
the second 8Kb extent that we successfully fsynced:
wrote 8192/8192 bytes at offset 0
XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
wrote 8192/8192 bytes at offset 8192
XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
File content after log replay: 0000000 aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa aa
* 0020000
So fix this by skipping the fsync only if we're doing a full sync and
if the inode's last_trans is <= fs_info->last_trans_committed, or if
the inode is already in the log. Also remove setting the inode's
last_trans in btrfs_file_write_iter since it's useless/unreliable.
Also because btrfs_file_write_iter no longer sets inode->last_trans to
fs_info->generation + 1, don't set last_trans to 0 if we bail out and don't
bail out if last_trans is 0, otherwise something as simple as the following
example wouldn't log the second write on the last fsync:
1. write to file
2. fsync file
3. fsync file
|--> btrfs_inode_in_log() returns true and it set last_trans to 0
4. write to file
|--> btrfs_file_write_iter() no longers sets last_trans, so it
remained with a value of 0
5. fsync
|--> inode->last_trans == 0, so it bails out without logging the
second write
A test case for xfstests will be sent soon.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Chris Mason <clm@fb.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A block-local variable stores error code but btrfs_get_blocks_direct may
not return it in the end as there's a ret defined in the function scope.
Fixes: d187663ef24c ("Btrfs: lock extents as we map them in DIO") Signed-off-by: David Sterba <dsterba@suse.cz> Signed-off-by: Chris Mason <clm@fb.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We can have multiple fsync operations against the same file during the
same transaction and they can collect the same ordered extents while they
don't complete (still accessible from the inode's ordered tree). If this
happens, those ordered extents will never get their reference counts
decremented to 0, leading to memory leaks and inode leaks (an iput for an
ordered extent's inode is scheduled only when the ordered extent's refcount
drops to 0). The following sequence diagram explains this race:
mutex_lock(inode->i_mutex)
btrfs_sync_log()
btrfs_wait_logged_extents()
--> list_del_init(&ordered->log_list)
btrfs_log_inode()
btrfs_get_logged_extents()
--> Adds ordered extent X
to logged_list because
at this point:
list_empty(&ordered->log_list)
&& test_bit(BTRFS_ORDERED_LOGGED,
&ordered->flags) == 0
--> Increments ordered extent
X's refcount
--> check if ordered extent's io is
finished or not, start it if
necessary and wait for it to finish
--> sets bit BTRFS_ORDERED_LOGGED
on ordered extent X's flags
and adds it to trans->ordered
btrfs_sync_log() finishes
btrfs_sync_log()
btrfs_wait_logged_extents()
--> Sees ordered extent X has the
bit BTRFS_ORDERED_LOGGED set in
its flags
--> X's refcount is untouched
btrfs_sync_log() finishes
btrfs_sync_file() finishes
btrfs_commit_transaction()
--> called by transaction kthread for e.g.
btrfs_wait_pending_ordered()
--> waits for ordered extent X to
complete
--> decrements ordered extent X's
refcount by 1 only, corresponding
to the increment done by the fsync
task ran by CPU 1
In the scenario of the above diagram, after the transaction commit,
the ordered extent will remain with a refcount of 1 forever, leaking
the ordered extent structure and preventing the i_count of its inode
from ever decreasing to 0, since the delayed iput is scheduled only
when the ordered extent's refcount drops to 0, preventing the inode
from ever being evicted by the VFS.
Fix this by using the flag BTRFS_ORDERED_LOGGED differently. Use it to
mean that an ordered extent is already being processed by an fsync call,
which will attach it to the current transaction, preventing it from being
collected by subsequent fsync operations against the same inode.
This race was introduced with the following change (added in 3.19 and
backported to stable 3.18 and 3.17):
Set the internal device state to to disabled after hardware reset in stop flow.
This will cover cases when driver was not brought to disabled state because of
an error and in stop flow we wish not to retry the reset.
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com> Signed-off-by: Tomas Winkler <tomas.winkler@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It incorrectly assumes that the level of indirection is not needed
which is not true(probably because the driver incorrectly allocates
sizeof(*client) instead of sizeof(*data) via devm_iio_device_alloc).
If you look at the code of the probe function(see below) it is easy to
see that what is being stored in the private memory of the IIO device
instance is not a copy of a 'struct i2c_client' but a pointer to an
instance passed as an argument to the probe function.
struct i2c_client **data;
int ret;
< Some code skipped >
indio_dev = devm_iio_device_alloc(&client->dev, sizeof(*client));
if (!indio_dev)
return -ENOMEM;
data = iio_priv(indio_dev);
*data = client;
Without reverting this change any read of a raw value of this sensor
leads to a kernel oops due to a NULL pointer de-reference on my
hardware setup.
Signed-off-by: Jonathan Cameron <jic23@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since only a pointer to struct i2c_client is stored in a private area
of IIO device created by the driver there's no need to allocate
sizeof(struct i2c_client) worth of storage.
Pushed to stable as this is linked to the revert patch previously.
Without this followup the original patch looks sensible.
Using the touchscreen while running buffered capture results in the
buffer reporting lots of wrong values, often just zeros. This is because
we push readings to the buffer every time a touchscreen interrupt
arrives, including when the buffer's own conversions have not yet
finished. So let's only push to the buffer when its conversions are
ready.
Signed-off-by: Kristina Martšenko <kristina.martsenko@gmail.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Jonathan Cameron <jic23@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reading a channel through sysfs, or starting a buffered capture, can
occasionally turn off the touchscreen.
This is because the read_raw() and buffer preenable()/postdisable()
callbacks unschedule current conversions on all channels. If a delay
channel happens to schedule a touchscreen conversion at the same time,
the conversion gets cancelled and the touchscreen sequence stops.
This is probably related to this note from the reference manual:
"If a delay group schedules channels to be sampled and a manual
write to the schedule field in CTRL0 occurs while the block is
discarding samples, the LRADC will switch to the new schedule
and will not sample the channels that were previously scheduled.
The time window for this to happen is very small and lasts only
while the LRADC is discarding samples."
So make the callbacks only unschedule conversions for the channels they
use. This means channel 0 for read_raw() and channels 0-5 for the buffer
(if the touchscreen is enabled). Since the touchscreen uses different
channels (6 and 7), it no longer gets turned off.
This is tested and fixes the issue on i.MX28, but hasn't been tested on
i.MX23.
Signed-off-by: Kristina Martšenko <kristina.martsenko@gmail.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Jonathan Cameron <jic23@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reading a channel through sysfs, or starting a buffered capture, will
currently turn off the touchscreen. This is because the read_raw() and
buffer preenable()/postdisable() callbacks disable interrupts for all
LRADC channels, including those the touchscreen uses.
So make the callbacks only disable interrupts for the channels they use.
This means channel 0 for read_raw() and channels 0-5 for the buffer (if
the touchscreen is enabled). Since the touchscreen uses different
channels (6 and 7), it no longer gets turned off.
Note that only i.MX28 is affected by this issue, i.MX23 should be fine.
Signed-off-by: Kristina Martšenko <kristina.martsenko@gmail.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Jonathan Cameron <jic23@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The touchscreen was initially designed [1] to map all of its physical
channels to one virtual channel, leaving buffered capture to use the
remaining 7 virtual channels. When the touchscreen was reimplemented
[2], it was made to use four virtual channels, which overlap and
conflict with the channels the buffer uses.
As a result, when the buffer is enabled, the touchscreen's virtual
channels are remapped to whichever physical channels the buffer was
configured with, causing the touchscreen to read those instead of the
touch measurement channels. Effectively the touchscreen stops working.
So here we separate the channels again, giving the touchscreen 2 virtual
channels and the buffer 6. We can't give the touchscreen just 1 channel
as before, as the current pressure calculation requires 2 channels to be
read at the same time.
This makes the touchscreen continue to work during buffered capture. It
has been tested on i.MX28, but not on i.MX23.
[1] 06ddd353f5c8 ("iio: mxs: Implement support for touchscreen")
[2] dee05308f602 ("Staging/iio/adc/touchscreen/MXS: add interrupt driven
touch detection")
Signed-off-by: Kristina Martšenko <kristina.martsenko@gmail.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Jonathan Cameron <jic23@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The intention is obviously to sign-extend a 12 bit quantity. But
because of C's promotion rules, the assignment is equivalent to "val16
&= 0xfff;". Use the proper API for this.
Since commit c8231a9af8147f8a ("iio: mxs-lradc: compute temperature
from channel 8 and 9") with the removal of adc channel 9 there is
no 1-1 mapping in the channel spec.
All hwmon channel values above 9 are accessible via there index minus
one. So add a hidden iio channel 9 to fix this issue.
Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com> Acked-by: Alexandre Belloni <alexandre.belloni@free-electrons.com> Reviewed-by: Marek Vasut <marex@denx.de> Signed-off-by: Jonathan Cameron <jic23@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
f31a9f7c7169 ("x86/xsaves: Use xsaves/xrstors to save and restore xsave area")
introduced alternative instructions for XSAVES/XRSTORS and commit:
adb9d526e982 ("x86/xsaves: Add xsaves and xrstors support for booting time")
added support for the XSAVES/XRSTORS instructions at boot time.
Unfortunately both failed to properly protect them against faulting:
The 'xstate_fault' macro will use the closest label named '1'
backward and that ends up in the .altinstr_replacement section
rather than in .text. This means that the kernel will never find
in the __ex_table the .text address where this instruction might
fault, leading to serious problems if userspace manages to
trigger the fault.
Signed-off-by: Quentin Casasnovas <quentin.casasnovas@oracle.com> Signed-off-by: Jamie Iles <jamie.iles@oracle.com>
[ Improved the changelog, fixed some whitespace noise. ] Acked-by: Borislav Petkov <bp@alien8.de> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Cc: Allan Xavier <mr.a.xavier@gmail.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Thomas Gleixner <tglx@linutronix.de> Fixes: adb9d526e982 ("x86/xsaves: Add xsaves and xrstors support for booting time") Fixes: f31a9f7c7169 ("x86/xsaves: Use xsaves/xrstors to save and restore xsave area") Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
'ret_from_fork' checks TIF_IA32 to determine whether 'pt_regs' and
the related state make sense for 'ret_from_sys_call'. This is
entirely the wrong check. TS_COMPAT would make a little more
sense, but there's really no point in keeping this optimization
at all.
This fixes a return to the wrong user CS if we came from int
0x80 in a 64-bit task.
Signed-off-by: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/4710be56d76ef994ddf59087aad98c000fbab9a4.1424989793.git.luto@amacapital.net
[ Backported from tip:x86/asm. ] Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds a check to sbc_parse_cdb() in order to detect when
an LBA + sector vs. end-of-device calculation wraps when the LBA is
sufficently large enough (eg: 0xFFFFFFFFFFFFFFFF).
Cc: Martin Petersen <martin.petersen@oracle.com> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch addresses the original PR_APTPL_BUF_LEN = 8k limitiation
for write-out of PR APTPL metadata that Martin has recently been
running into.
It changes core_scsi3_update_and_write_aptpl() to use vzalloc'ed
memory instead of kzalloc, and increases the default hardcoded
length to 256k.
It also adds logic in core_scsi3_update_and_write_aptpl() to double
the original length upon core_scsi3_update_aptpl_buf() failure, and
retries until the vzalloc'ed buffer is large enough to accommodate
the outgoing APTPL metadata.
Reported-by: Martin Svec <martin.svec@zoner.cz> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The efficient frequency (RPe) should stay in the range
RPn <= RPe <= RP0. The pcode clamps the returned value
internally on Broadwell but not on Haswell.
Fix for missing range check in
commit 93ee29203f506582cca2bcec5f05041526d9ab0a
Author: Tom O'Rourke <Tom.O'Rourke@intel.com>
Date: Wed Nov 19 14:21:52 2014 -0800
drm/i915: Use efficient frequency for HSW/BDW
Reference: http://lists.freedesktop.org/archives/intel-gfx/2015-February/059802.html Reported-by: Michael Auchter <a@phire.org> Suggested-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Tom O'Rourke <Tom.O'Rourke@intel.com> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
As per the specififcation, the SB_DevFn is the PCI_DEVFN of the target
device and not the source. So PCI_DEVFN(2,0) is not correct. Further the
port ID should be enough to identify devices unless they are MFD. The
SB_DevFn was intended to remove ambiguity in case of these MFD devices.
For non MFD devices the recommendation for the target device IP was to
ignore these fields, but not all of them followed the recommendation.
Some like CCK ignore these fields and hence PCI_DEVFN(2, 0) works and so
does PCI_DEVFN(0, 0) as it works for DPIO. The issue came to light because
of GPIONC which was not getting programmed correctly with PCI_DEVFN(2, 0).
It turned out that this did not follow the recommendation and expected 0
in this field.
In general the recommendation is to use SB_DevFn as PCI_DEVFN(0, 0) for
all devices except target PCI devices.
Signed-off-by: Shobhit Kumar <shobhit.kumar@intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It's possible for invalidate_range_start mmu notifier callback to race
against userptr object release. If the gem object was released prior to
obtaining the spinlock in invalidate_range_start we're hitting null
pointer dereference.
Testcase: igt/gem_userptr_blits/stress-mm-invalidate-close
Testcase: igt/gem_userptr_blits/stress-mm-invalidate-close-overlap Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Michał Winiarski <michal.winiarski@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[Jani: added code comment suggested by Chris] Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Nothing in Bspec seems to indicate that we actually needs this, and it
looks like can't work since by this point the pipe is off and so
vblanks won't really happen any more.
Note that Bspec mentions that it takes a vblank for this bit to
change, but _only_ when enabling.
Dropping this code quenches an annoying backtrace introduced by the
more anal checking since
Note: This fixes the fallout from the above commit, but does not address
the shortcomings of the IBX transcoder select workaround implementation
discussed during review [1].
[1] http://mid.gmane.org/87y4o7usxf.fsf@intel.com
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86095 Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
but in reality it undercovered a much older coherency bug. The issue that
boosting the GPU frequency on the BCS ring was masking was that we could
wake the CPU up after completion of a BCS batch and inspect memory prior
to the write cache being fully evicted. In order to serialise the
breadcrumb interrupt (and so ensure that the CPU's view of memory is
coherent) we need to perform a post-sync operation in the MI_FLUSH_DW.
v2: Fix all the MI_FLUSH_DW (bsd plus the duplication in execlists).
Also fix the invalidate_domains mask in gen8_emit_flush() for ring !=
VCS.
Testcase: gpuX-rcs-gpu-read-after-write Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Daniel Vetter <daniel@ffwll.ch> Signed-off-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Historically, !__GFP_FS allocations were not allowed to invoke the OOM
killer once reclaim had failed, but nevertheless kept looping in the
allocator.
Commit 9879de7373fc ("mm: page_alloc: embed OOM killing naturally into
allocation slowpath"), which should have been a simple cleanup patch,
accidentally changed the behavior to aborting the allocation at that
point. This creates problems with filesystem callers (?) that currently
rely on the allocator waiting for other tasks to intervene.
Revert the behavior as it shouldn't have been changed as part of a
cleanup patch.
Fixes: 9879de7373fc ("mm: page_alloc: embed OOM killing naturally into allocation slowpath") Signed-off-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Michal Hocko <mhocko@suse.cz> Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Dave Chinner <david@fromorbit.com> Acked-by: David Rientjes <rientjes@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Mel Gorman <mgorman@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Maxime reported the following memory leak regression due to commit dbc8358c7237 ("mm/nommu: use alloc_pages_exact() rather than its own
implementation").
On v3.19, I am facing a memory leak. Each time I run a command one page
is lost. Here an example with busybox's free command:
/ # free
total used free shared buffers cached
Mem: 7928 1972 5956 0 0 492
-/+ buffers/cache: 1480 6448
/ # free
total used free shared buffers cached
Mem: 7928 1976 5952 0 0 492
-/+ buffers/cache: 1484 6444
/ # free
total used free shared buffers cached
Mem: 7928 1980 5948 0 0 492
-/+ buffers/cache: 1488 6440
/ # free
total used free shared buffers cached
Mem: 7928 1984 5944 0 0 492
-/+ buffers/cache: 1492 6436
/ # free
total used free shared buffers cached
Mem: 7928 1988 5940 0 0 492
-/+ buffers/cache: 1496 6432
At some point, the system fails to sastisfy 256KB allocations:
This problem happens because we allocate ordered page through
__get_free_pages() in do_mmap_private() in some cases and we try to free
individual pages rather than ordered page in free_page_series(). In
this case, freeing pages whose refcount is not 0 won't be freed to the
page allocator so memory leak happens.
To fix the problem, this patch changes __get_free_pages() to
alloc_pages_exact() since alloc_pages_exact() returns
physically-contiguous pages but each pages are refcounted.
Fixes: dbc8358c7237 ("mm/nommu: use alloc_pages_exact() rather than its own implementation"). Reported-by: Maxime Coquelin <mcoquelin.stm32@gmail.com> Tested-by: Maxime Coquelin <mcoquelin.stm32@gmail.com> Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.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@linuxfoundation.org>
The vmstat interfaces are good at hiding negative counts (at least when
CONFIG_SMP); but if you peer behind the curtain, you find that
nr_isolated_anon and nr_isolated_file soon go negative, and grow ever
more negative: so they can absorb larger and larger numbers of isolated
pages, yet still appear to be zero.
I'm happy to avoid a congestion_wait() when too_many_isolated() myself;
but I guess it's there for a good reason, in which case we ought to get
too_many_isolated() working again.
The imbalance comes from isolate_migratepages()'s ISOLATE_ABORT case:
putback_movable_pages() decrements the NR_ISOLATED counts, but we forgot
to call acct_isolated() to increment them.
It is possible that the bug whcih this patch fixes could cause OOM kills
when the system still has a lot of reclaimable page cache.
Fixes: edc2ca612496 ("mm, compaction: move pageblock checks up from isolate_migratepages_range()") Signed-off-by: Hugh Dickins <hughd@google.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Joonsoo Kim <iamjoonsoo.kim@lge.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@linuxfoundation.org>
A race condition starts to be visible in recent mmotm, where a PG_hwpoison
flag is set on a migration source page *before* it's back in buddy page
poo= l.
This is problematic because no page flag is supposed to be set when
freeing (see __free_one_page().) So the user-visible effect of this race
is that it could trigger the BUG_ON() when soft-offlining is called.
The root cause is that we call lru_add_drain_all() to make sure that the
page is in buddy, but that doesn't work because this function just
schedule= s a work item and doesn't wait its completion.
drain_all_pages() does drainin= g directly, so simply dropping
lru_add_drain_all() solves this problem.
Fixes: f15bdfa802bf ("mm/memory-failure.c: fix memory leak in successful soft offlining") Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Tony Luck <tony.luck@intel.com> Cc: Chen Gong <gong.chen@linux.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@linuxfoundation.org>
For whatever reason, generic_access_phys() only remaps one page, but
actually allows to access arbitrary size. It's quite easy to trigger
large reads, like printing out large structure with gdb, which leads to a
crash. Fix it by remapping correct size.
What we want to check here is whether there is highorder freepage in buddy
list of other migratetype in order to steal it without fragmentation.
But, current code just checks cc->order which means allocation request
order. So, this is wrong.
Without this fix, non-movable synchronous compaction below pageblock order
would not stopped until compaction is complete, because migratetype of
most pageblocks are movable and high order freepage made by compaction is
usually on movable type buddy list.
There is some report related to this bug. See below link.
Although the issued system still has load spike comes from compaction,
this makes that system completely stable and responsive according to his
report.
stress-highalloc test in mmtests with non movable order 7 allocation
doesn't show any notable difference in allocation success rate, but, it
shows more compaction success rate.
I noticed that "allowed" can easily overflow by falling below 0, because
(total_vm / 32) can be larger than "allowed". The problem occurs in
OVERCOMMIT_NONE mode.
In this case, a huge allocation can success and overcommit the system
(despite OVERCOMMIT_NONE mode). All subsequent allocations will fall
(system-wide), so system become unusable.
The problem was masked out by commit c9b1d0981fcc
("mm: limit growth of 3% hardcoded other user reserve"),
but it's easy to reproduce it on older kernels:
1) set overcommit_memory sysctl to 2
2) mmap() large file multiple times (with VM_SHARED flag)
3) try to malloc() large amount of memory
It also can be reproduced on newer kernels, but miss-configured
sysctl_user_reserve_kbytes is required.
Fix this issue by switching to signed arithmetic here.
Signed-off-by: Roman Gushchin <klamm@yandex-team.ru> Cc: Andrew Shewmaker <agshew@gmail.com> Cc: Rik van Riel <riel@redhat.com> Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru> 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@linuxfoundation.org>
I noticed, that "allowed" can easily overflow by falling below 0,
because (total_vm / 32) can be larger than "allowed". The problem
occurs in OVERCOMMIT_NONE mode.
In this case, a huge allocation can success and overcommit the system
(despite OVERCOMMIT_NONE mode). All subsequent allocations will fall
(system-wide), so system become unusable.
The problem was masked out by commit c9b1d0981fcc
("mm: limit growth of 3% hardcoded other user reserve"),
but it's easy to reproduce it on older kernels:
1) set overcommit_memory sysctl to 2
2) mmap() large file multiple times (with VM_SHARED flag)
3) try to malloc() large amount of memory
It also can be reproduced on newer kernels, but miss-configured
sysctl_user_reserve_kbytes is required.
Fix this issue by switching to signed arithmetic here.
[akpm@linux-foundation.org: use min_t] Signed-off-by: Roman Gushchin <klamm@yandex-team.ru> Cc: Andrew Shewmaker <agshew@gmail.com> Cc: Rik van Riel <riel@redhat.com> Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru> Reviewed-by: Michal Hocko <mhocko@suse.cz> 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@linuxfoundation.org>
When studying page stealing, I noticed some weird looking decisions in
try_to_steal_freepages(). The first I assume is a bug (Patch 1), the
following two patches were driven by evaluation.
Testing was done with stress-highalloc of mmtests, using the
mm_page_alloc_extfrag tracepoint and postprocessing to get counts of how
often page stealing occurs for individual migratetypes, and what
migratetypes are used for fallbacks. Arguably, the worst case of page
stealing is when UNMOVABLE allocation steals from MOVABLE pageblock.
RECLAIMABLE allocation stealing from MOVABLE allocation is also not ideal,
so the goal is to minimize these two cases.
The evaluation of v2 wasn't always clear win and Joonsoo questioned the
results. Here I used different baseline which includes RFC compaction
improvements from [1]. I found that the compaction improvements reduce
variability of stress-highalloc, so there's less noise in the data.
First, let's look at stress-highalloc configured to do sync compaction,
and how these patches reduce page stealing events during the test. First
column is after fresh reboot, other two are reiterations of test without
reboot. That was all accumulater over 5 re-iterations (so the benchmark
was run 5x3 times with 5 fresh restarts).
With Patch 1:
3.19-rc4 3.19-rc4 3.19-rc4
6-nothp-1 6-nothp-2 6-nothp-3
Page alloc extfrag event 1183495498775239774860
Extfrag fragmenting 1183399398768809774245
Extfrag fragmenting for unmovable 7342 16129 11712
Extfrag fragmenting unmovable placed with movable 4191 10547 6270
Extfrag fragmenting for reclaimable 373 1130 923
Extfrag fragmenting reclaimable placed with movable 302 906 738
Extfrag fragmenting for movable 1182627898596219761610
With Patch 2:
3.19-rc4 3.19-rc4 3.19-rc4
7-nothp-1 7-nothp-2 7-nothp-3
Page alloc extfrag event 472599036687933807436
Extfrag fragmenting 472510436682523806898
Extfrag fragmenting for unmovable 6678 7974 7281
Extfrag fragmenting unmovable placed with movable 2051 3829 4017
Extfrag fragmenting for reclaimable 429 1208 1278
Extfrag fragmenting reclaimable placed with movable 369 976 1034
Extfrag fragmenting for movable 471799736590703798339
With Patch 3:
3.19-rc4 3.19-rc4 3.19-rc4
8-nothp-1 8-nothp-2 8-nothp-3
Page alloc extfrag event 501618347001423850633
Extfrag fragmenting 501532546996133850072
Extfrag fragmenting for unmovable 1312 3154 3088
Extfrag fragmenting unmovable placed with movable 1115 2777 2714
Extfrag fragmenting for reclaimable 437 1193 1097
Extfrag fragmenting reclaimable placed with movable 330 969 879
Extfrag fragmenting for movable 501357646952663845887
In v2 we've seen apparent regression with Patch 1 for unmovable events,
this is now gone, suggesting it was indeed noise. Here, each patch
improves the situation for unmovable events. Reclaimable is improved by
patch 1 and then either the same modulo noise, or perhaps sligtly worse -
a small price for unmovable improvements, IMHO. The number of movable
allocations falling back to other migratetypes is most noisy, but it's
reduced to half at Patch 2 nevertheless. These are least critical as
compaction can move them around.
If we look at success rates, the patches don't affect them, that didn't change.
While there's no improvement here, I consider reduced fragmentation events
to be worth on its own. Patch 2 also seems to reduce scanning for free
pages, and migrations in compaction, suggesting it has somewhat less work
to do:
As with v2, /proc/pagetypeinfo appears unaffected with respect to numbers
of unmovable and reclaimable pageblocks.
Configuring the benchmark to allocate like THP page fault (i.e. no sync
compaction) gives much noisier results for iterations 2 and 3 after
reboot. This is not so surprising given how [1] offers lower improvements
in this scenario due to less restarts after deferred compaction which
would change compaction pivot.
When __rmqueue_fallback() is called to allocate a page of order X, it will
find a page of order Y >= X of a fallback migratetype, which is different
from the desired migratetype. With the help of try_to_steal_freepages(),
it may change the migratetype (to the desired one) also of:
1) all currently free pages in the pageblock containing the fallback page
2) the fallback pageblock itself
3) buddy pages created by splitting the fallback page (when Y > X)
These decisions take the order Y into account, as well as the desired
migratetype, with the goal of preventing multiple fallback allocations
that could e.g. distribute UNMOVABLE allocations among multiple
pageblocks.
Originally, decision for 1) has implied the decision for 3). Commit 47118af076f6 ("mm: mmzone: MIGRATE_CMA migration type added") changed that
(probably unintentionally) so that the buddy pages in case 3) are always
changed to the desired migratetype, except for CMA pageblocks.
Commit fef903efcf0c ("mm/page_allo.c: restructure free-page stealing code
and fix a bug") did some refactoring and added a comment that the case of
3) is intended. Commit 0cbef29a7821 ("mm: __rmqueue_fallback() should
respect pageblock type") removed the comment and tried to restore the
original behavior where 1) implies 3), but due to the previous
refactoring, the result is instead that only 2) implies 3) - and the
conditions for 2) are less frequently met than conditions for 1). This
may increase fragmentation in situations where the code decides to steal
all free pages from the pageblock (case 1)), but then gives back the buddy
pages produced by splitting.
This patch restores the original intended logic where 1) implies 3).
During testing with stress-highalloc from mmtests, this has shown to
decrease the number of events where UNMOVABLE and RECLAIMABLE allocations
steal from MOVABLE pageblocks, which can lead to permanent fragmentation.
In some cases it has increased the number of events when MOVABLE
allocations steal from UNMOVABLE or RECLAIMABLE pageblocks, but these are
fixable by sync compaction and thus less harmful.
Note that evaluation has shown that the behavior introduced by 47118af076f6 for buddy pages in case 3) is actually even better than the
original logic, so the following patch will introduce it properly once
again. For stable backports of this patch it makes thus sense to only fix
versions containing 0cbef29a7821.
[iamjoonsoo.kim@lge.com: tracepoint fix] Signed-off-by: Vlastimil Babka <vbabka@suse.cz> Acked-by: Mel Gorman <mgorman@suse.de> Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com> Acked-by: Minchan Kim <minchan@kernel.org> Cc: David Rientjes <rientjes@google.com> Cc: Rik van Riel <riel@redhat.com> Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Michal Hocko <mhocko@suse.cz> Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.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@linuxfoundation.org>
Commit ed4d4902ebdd ("mm, hugetlb: remove hugetlb_zero and
hugetlb_infinity") replaced 'unsigned long hugetlb_zero' with 'int zero'
leading to out-of-bounds access in proc_doulongvec_minmax(). Use
'.extra1 = NULL' instead of '.extra1 = &zero'. Passing NULL is
equivalent to passing minimal value, which is 0 for unsigned types.
If __unmap_hugepage_range() tries to unmap the address range over which
hugepage migration is on the way, we get the wrong page because pte_page()
doesn't work for migration entries. This patch simply clears the pte for
migration entries as we do for hwpoison entries.
Fixes: 290408d4a2 ("hugetlb: hugepage migration core") Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Hugh Dickins <hughd@google.com> Cc: James Hogan <james.hogan@imgtec.com> Cc: David Rientjes <rientjes@google.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@suse.cz> Cc: Rik van Riel <riel@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Luiz Capitulino <lcapitulino@redhat.com> Cc: Nishanth Aravamudan <nacc@linux.vnet.ibm.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: Steve Capper <steve.capper@linaro.org> 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@linuxfoundation.org>
There is a race condition between hugepage migration and
change_protection(), where hugetlb_change_protection() doesn't care about
migration entries and wrongly overwrites them. That causes unexpected
results like kernel crash. HWPoison entries also can cause the same
problem.
This patch adds is_hugetlb_entry_(migration|hwpoisoned) check in this
function to do proper actions.
Fixes: 290408d4a2 ("hugetlb: hugepage migration core") Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Hugh Dickins <hughd@google.com> Cc: James Hogan <james.hogan@imgtec.com> Cc: David Rientjes <rientjes@google.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@suse.cz> Cc: Rik van Riel <riel@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Luiz Capitulino <lcapitulino@redhat.com> Cc: Nishanth Aravamudan <nacc@linux.vnet.ibm.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: Steve Capper <steve.capper@linaro.org> 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@linuxfoundation.org>
When running the test which causes the race as shown in the previous patch,
we can hit the BUG "get_page() on refcount 0 page" in hugetlb_fault().
This race happens when pte turns into migration entry just after the first
check of is_hugetlb_entry_migration() in hugetlb_fault() passed with false.
To fix this, we need to check pte_present() again after huge_ptep_get().
This patch also reorders taking ptl and doing pte_page(), because
pte_page() should be done in ptl. Due to this reordering, we need use
trylock_page() in page != pagecache_page case to respect locking order.
Fixes: 66aebce747ea ("hugetlb: fix race condition in hugetlb_fault()") Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Hugh Dickins <hughd@google.com> Cc: James Hogan <james.hogan@imgtec.com> Cc: David Rientjes <rientjes@google.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@suse.cz> Cc: Rik van Riel <riel@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Luiz Capitulino <lcapitulino@redhat.com> Cc: Nishanth Aravamudan <nacc@linux.vnet.ibm.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: Steve Capper <steve.capper@linaro.org> 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@linuxfoundation.org>
Currently the list is traversed using rcu variant. That is not correct
since dev_set_mac_address can be called which eventually calls
rtmsg_ifinfo_build_skb and there, skb allocation can sleep. So fix this
by remove the rcu usage here.
Fixes: 3d249d4ca7 "net: introduce ethernet teaming device" Signed-off-by: Jiri Pirko <jiri@resnulli.us> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
1. For an IPv4 ping socket, ping_check_bind_addr does not check
the family of the socket address that's passed in. Instead,
make it behave like inet_bind, which enforces either that the
address family is AF_INET, or that the family is AF_UNSPEC and
the address is 0.0.0.0.
2. For an IPv6 ping socket, ping_check_bind_addr returns EINVAL
if the socket family is not AF_INET6. Return EAFNOSUPPORT
instead, for consistency with inet6_bind.
3. Make ping_v4_sendmsg and ping_v6_sendmsg return EAFNOSUPPORT
instead of EINVAL if an incorrect socket address structure is
passed in.
4. Make IPv6 ping sockets be IPv6-only. The code does not support
IPv4, and it cannot easily be made to support IPv4 because
the protocol numbers for ICMP and ICMPv6 are different. This
makes connect(::ffff:192.0.2.1) fail with EAFNOSUPPORT instead
of making the socket unusable.
Among other things, this fixes an oops that can be triggered by:
Change-Id: If06ca86d9f1e4593c0d6df174caca3487c57a241 Signed-off-by: Lorenzo Colitti <lorenzo@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>