- 06 Dec, 2016 40 commits
-
-
Jaehoon Chung authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 45c7a490 upstream. platform_get_resource can be returned the NULL pointer. Then regs->start should be referred to NULL Pointer. devm_ioremap_resource() checks whether res is NULL or not. Signed-off-by:
Jaehoon Chung <jh80.chung@samsung.com> Reviewed-by:
Shawn Lin <shawn.lin@rock-chips.com> Signed-off-by:
Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by:
Luis Henriques <luis.henriques@canonical.com>
-
Ching Huang authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 2bf7dc84 upstream. The arcmsr driver failed to pass SYNCHRONIZE CACHE to controller firmware. Depending on how drive caches are handled internally by controller firmware this could potentially lead to data integrity problems. Ensure that cache flushes are passed to the controller. [mkp: applied by hand and removed unused vars] Signed-off-by:
Ching Huang <ching2048@areca.com.tw> Reported-by:
Tomas Henzl <thenzl@redhat.com> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by:
-
Ewan D. Milne authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 4d2b496f upstream. map_storep was not being vfree()'d in the module_exit call. Signed-off-by:
Ewan D. Milne <emilne@redhat.com> Reviewed-by:
Laurence Oberman <loberman@redhat.com> Signed-off-by:
-
Kashyap Desai authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 1e793f6f upstream. Commit 02b01e01 ("megaraid_sas: return sync cache call with success") modified the driver to successfully complete SYNCHRONIZE_CACHE commands without passing them to the controller. Disk drive caches are only explicitly managed by controller firmware when operating in RAID mode. So this commit effectively disabled writeback cache flushing for any drives used in JBOD mode, leading to data integrity failures. [mkp: clarified patch description] Fixes: 02b01e01Signed-off-by:
Kashyap Desai <kashyap.desai@broadcom.com> Signed-off-by:
Sumit Saxena <sumit.saxena@broadcom.com> Reviewed-by:
Hannes Reinecke <hare@suse.com> Reviewed-by:
-
Johannes Berg authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit ea720935 upstream. In mac80211, multicast A-MSDUs are accepted in many cases that they shouldn't be accepted in: * drop A-MSDUs with a multicast A1 (RA), as required by the spec in 9.11 (802.11-2012 version) * drop A-MSDUs with a 4-addr header, since the fourth address can't actually be useful for them; unless 4-address frame format is actually requested, even though the fourth address is still not useful in this case, but ignored Accepting the first case, in particular, is very problematic since it allows anyone else with possession of a GTK to send unicast frames encapsulated in a multicast A-MSDU, even when the AP has client isolation enabled. Signed-off-by:
Johannes Berg <johannes.berg@intel.com> Signed-off-by:
-
Stefan Richter authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit e9300a4b upstream. RFC 2734 defines the datagram_size field in fragment encapsulation headers thus: datagram_size: The encoded size of the entire IP datagram. The value of datagram_size [...] SHALL be one less than the value of Total Length in the datagram's IP header (see STD 5, RFC 791). Accordingly, the eth1394 driver of Linux 2.6.36 and older set and got this field with a -/+1 offset: ether1394_tx() /* transmit */ ether1394_encapsulate_prep() hdr->ff.dg_size = dg_size - 1; ether1394_data_handler() /* receive */ if (hdr->common.lf == ETH1394_HDR_LF_FF) dg_size = hdr->ff.dg_size + 1; else dg_size = hdr->sf.dg_size + 1; Likewise, I observe OS X 10.4 and Windows XP Pro SP3 to transmit 1500 byte sized datagrams in fragments with datagram_size=1499 if link fragmentation is required. Only firewire-net sets and gets datagram_size without this offset. The result is lacking interoperability of firewire-net with OS X, Windows XP, and presumably Linux' eth1394. (I did not test with the latter.) For example, FTP data transfers to a Linux firewire-net box with max_rec smaller than the 1500 bytes MTU - from OS X fail entirely, - from Win XP start out with a bunch of fragmented datagrams which time out, then continue with unfragmented datagrams because Win XP temporarily reduces the MTU to 576 bytes. So let's fix firewire-net's datagram_size accessors. Note that firewire-net thereby loses interoperability with unpatched firewire-net, but only if link fragmentation is employed. (This happens with large broadcast datagrams, and with large datagrams on several FireWire CardBus cards with smaller max_rec than equivalent PCI cards, and it can be worked around by setting a small enough MTU.) Signed-off-by:
Stefan Richter <stefanr@s5r6.in-berlin.de> Signed-off-by:
-
Stefan Richter authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 667121ac upstream. The IP-over-1394 driver firewire-net lacked input validation when handling incoming fragmented datagrams. A maliciously formed fragment with a respectively large datagram_offset would cause a memcpy past the datagram buffer. So, drop any packets carrying a fragment with offset + length larger than datagram_size. In addition, ensure that - GASP header, unfragmented encapsulation header, or fragment encapsulation header actually exists before we access it, - the encapsulated datagram or fragment is of nonzero size. Reported-by:
Eyal Itkin <eyal.itkin@gmail.com> Reviewed-by:
-
Patrick Scheuring authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit da25311c upstream. The Schenker XMG C504 is a rebranded Gigabyte P35 v2 laptop. Therefore it also needs a keyboard reset to detect the Elantech touchpad. Otherwise the touchpad appears to be dead. With this patch the touchpad is detected: $ dmesg | grep -E "(i8042|Elantech|elantech)" [ 2.675399] i8042: PNP: PS/2 Controller [PNP0303:PS2K,PNP0f13:PS2M] at 0x60,0x64 irq 1,12 [ 2.680372] i8042: Attempting to reset device connected to KBD port [ 2.789037] serio: i8042 KBD port at 0x60,0x64 irq 1 [ 2.791586] serio: i8042 AUX port at 0x60,0x64 irq 12 [ 2.813840] input: AT Translated Set 2 keyboard as /devices/platform/i8042/serio0/input/input4 [ 3.811431] psmouse serio1: elantech: assuming hardware version 4 (with firmware version 0x361f0e) [ 3.825424] psmouse serio1: elantech: Synaptics capabilities query result 0x00, 0x15, 0x0f. [ 3.839424] psmouse serio1: elantech: Elan sample query result 03, 58, 74 [ 3.911349] input: ETPS/2 Elantech Touchpad as /devices/platform/i8042/serio1/input/input6 Signed-off-by:
Patrick Scheuring <patrick.scheuring.dev@gmail.com> Signed-off-by:
Dmitry Torokhov <dmitry.torokhov@gmail.com> Signed-off-by:
-
Heinz Mauelshagen authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit dcb2ff56 upstream. If a default leg has failed, any read will cause a new operational default leg to be selected and the read is resubmitted. But until now the read will return failure even though it was successful due to resubmission. The reason for this is bio->bi_error was not being cleared before resubmitting the bio. Fix by clearing bio->bi_error before resubmission. Fixes: 4246a0b6 ("block: add a bi_error field to struct bio") Signed-off-by:
Heinz Mauelshagen <heinzm@redhat.com> Signed-off-by:
Mike Snitzer <snitzer@redhat.com> Signed-off-by:
-
Matt Redfearn authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 34563769 upstream. Commit c6017e79 ("virtio: console: add locks around buffer removal in port unplug path") added locking around the freeing of buffers in the vq. However, when free_buf() is called with can_sleep = true and rproc is enabled, it calls dma_free_coherent() directly, requiring interrupts to be enabled. Currently a WARNING is triggered due to the spin locking around free_buf, with a call stack like this: WARNING: CPU: 3 PID: 121 at ./include/linux/dma-mapping.h:433 free_buf+0x1a8/0x288 Call Trace: [<8040c538>] show_stack+0x74/0xc0 [<80757240>] dump_stack+0xd0/0x110 [<80430d98>] __warn+0xfc/0x130 [<80430ee0>] warn_slowpath_null+0x2c/0x3c [<807e7c6c>] free_buf+0x1a8/0x288 [<807ea590>] remove_port_data+0x50/0xac [<807ea6a0>] unplug_port+0xb4/0x1bc [<807ea858>] virtcons_remove+0xb0/0xfc [<807b6734>] virtio_dev_remove+0x58/0xc0 [<807f918c>] __device_release_driver+0xac/0x134 [<807f924c>] device_release_driver+0x38/0x50 [<807f7edc>] bus_remove_device+0xfc/0x130 [<807f4b74>] device_del+0x17c/0x21c [<807f4c38>] device_unregister+0x24/0x38 [<807b6b50>] unregister_virtio_device+0x28/0x44 Fix this by restructuring the loops to allow the locks to only be taken where it is necessary to protect the vqs, and release it while the buffer is being freed. Fixes: c6017e79 ("virtio: console: add locks around buffer removal in port unplug path") Signed-off-by:
Matt Redfearn <matt.redfearn@imgtec.com> Signed-off-by:
Michael S. Tsirkin <mst@redhat.com> Signed-off-by:
-
Ladi Prosek authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 0ea1e4a6 upstream. According to the spec, if the VIRTIO_RING_F_EVENT_IDX feature bit is negotiated the driver MUST set flags to 0. Not dirtying the available ring in virtqueue_disable_cb also has a minor positive performance impact, improving L1 dcache load missed by ~0.5% in vring_bench. Writes to the used event field (vring_used_event) are still unconditional. Cc: Michael S. Tsirkin <mst@redhat.com> Signed-off-by:
Ladi Prosek <lprosek@redhat.com> Signed-off-by:
-
John David Anglin authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 6ed51832 upstream. We have one critical section in the syscall entry path in which we switch from the userspace stack to kernel stack. In the event of an external interrupt, the interrupt code distinguishes between those two states by analyzing the value of sr7. If sr7 is zero, it uses the kernel stack. Therefore it's important, that the value of sr7 is in sync with the currently enabled stack. This patch now disables interrupts while executing the critical section. This prevents the interrupt handler to possibly see an inconsistent state which in the worst case can lead to crashes. Interestingly, in the syscall exit path interrupts were already disabled in the critical section which switches back to the userspace stack. Signed-off-by:
John David Anglin <dave.anglin@bell.net> Signed-off-by:
Helge Deller <deller@gmx.de> Signed-off-by:
-
Miklos Szeredi authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 641089c1 upstream. Make sure the copied up file hits the disk before renaming to the final destination. If this is not done then the copy-up may corrupt the data in the file in case of a crash. Signed-off-by:
Miklos Szeredi <mszeredi@redhat.com> Signed-off-by:
-
James Hogan authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit ede5f3e7 upstream. The ERET instruction to return from exception is used for returning from exception level (Status.EXL) and error level (Status.ERL). If both bits are set however we should be returning from ERL first, as ERL can interrupt EXL, for example when an NMI is taken. KVM however checks EXL first. Fix the order of the checks to match the pseudocode in the instruction set manual. Fixes: e685c689 ("KVM/MIPS32: Privileged instruction/target branch emulation.") Signed-off-by:
James Hogan <james.hogan@imgtec.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: "Radim Krčmář" <rkrcmar@redhat.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: linux-mips@linux-mips.org Cc: kvm@vger.kernel.org Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com> Signed-off-by:
-
Ido Yariv authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit bd768e14 upstream. vcpu->arch.wbinvd_dirty_mask may still be used after freeing it, corrupting memory. For example, the following call trace may set a bit in an already freed cpu mask: kvm_arch_vcpu_load vcpu_load vmx_free_vcpu_nested vmx_free_vcpu kvm_arch_vcpu_free Fix this by deferring freeing of wbinvd_dirty_mask. Signed-off-by:
Ido Yariv <ido@wizery.com> Reviewed-by:
Radim Krčmář <rkrcmar@redhat.com> Signed-off-by:
-
Tahsin Erdogan authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit d09960b0 upstream. dm_old_request_fn() has paths that access md->io_barrier. The party destroying io_barrier should ensure that no future execution of dm_old_request_fn() is possible. Move io_barrier destruction to below blk_cleanup_queue() to ensure this and avoid a NULL pointer crash during request-based DM device shutdown. Signed-off-by:
Tahsin Erdogan <tahsin@google.com> Signed-off-by:
-
Johan Hovold authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit de24e0a1 upstream. The current tiocmget implementation would fail to report errors up the stack and instead leaked a few bits from the stack as a mask of modem-status flags. Fixes: 39a66b8d ("[PATCH] USB: CP2101 Add support for flow control") Signed-off-by:
Johan Hovold <johan@kernel.org> Signed-off-by:
-
Dmitry Vyukov authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 32b2921e upstream. Size of kmalloc() in vc_do_resize() is controlled by user. Too large kmalloc() size triggers WARNING message on console. Put a reasonable upper bound on terminal size to prevent WARNINGs. Signed-off-by:
Dmitry Vyukov <dvyukov@google.com> CC: David Rientjes <rientjes@google.com> Cc: One Thousand Gnomes <gnomes@lxorguk.ukuu.org.uk> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Jiri Slaby <jslaby@suse.com> Cc: Peter Hurley <peter@hurleysoftware.com> Cc: linux-kernel@vger.kernel.org Cc: syzkaller@googlegroups.com Signed-off-by:
-
Mathias Nyman authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 4c39135a upstream. xHC in Wildcatpoint-LP PCH is similar to LynxPoint-LP and need the same quirks to prevent machines from spurious restart while shutting them down. Reported-by:
Hasan Mahmood <hasan.mahm@gmail.com> Signed-off-by:
Mathias Nyman <mathias.nyman@linux.intel.com> Signed-off-by:
-
Scot Doyle authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 009e39ae upstream. When resizing a vt its selection may exceed the new size, resulting in an invalid memory access [1]. Clear the selection before resizing. [1] http://lkml.kernel.org/r/CACT4Y+acDTwy4umEvf5ROBGiRJNrxHN4Cn5szCXE5Jw-d1B=Xw@mail.gmail.comReported-and-tested-by:
Scot Doyle <lkml14@scotdoyle.com> Signed-off-by:
-
Linus Torvalds authored
BugLink: http://bugs.launchpad.net/bugs/1642572 Not upstream as it is not needed there. So a patch something like this might be a safe way to fix the potential infoleak in older kernels. THIS IS UNTESTED. It's a very obvious patch, though, so if it compiles it probably works. It just initializes the output variable with 0 in the inline asm description, instead of doing it in the exception handler. It will generate slightly worse code (a few unnecessary ALU operations), but it doesn't have any interactions with the exception handler implementation. Signed-off-by:
-
Bryan Paluch authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit ed6d6f8f upstream. Increase ohci watchout delay to 275 ms. Previous delay was 250 ms with 20 ms of slack, after removing slack time some ohci controllers don't respond in time. Logs from systems with controllers that have the issue would show "HcDoneHead not written back; disabled" Signed-off-by:
Bryan Paluch <bryanpaluch@gmail.com> Acked-by:
Alan Stern <stern@rowland.harvard.edu> Signed-off-by:
-
Mathias Nyman authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 7d3b016a upstream. USB2 host inititated resume, and system suspend bus resume need to use the same USB_RESUME_TIMEOUT as elsewhere. This resolves a device disconnect issue at system resume seen on Intel Braswell and Apollolake, but is in no way limited to those platforms. Signed-off-by:
-
Stefan Tauner authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit ca006f78 upstream. This adds support to ftdi_sio for the Infineon TriBoard TC2X7 engineering board for first-generation Aurix SoCs with Tricore CPUs. Mere addition of the device IDs does the job. Signed-off-by:
Stefan Tauner <stefan.tauner@technikum-wien.at> Signed-off-by:
-
Johan Hovold authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 126d26f6 upstream. Make sure we have at least one port before attempting to register a console. Currently, at least one driver binds to a "dummy" interface and requests zero ports for it. Should such an interface also lack endpoints, we get a NULL-deref during probe. Fixes: e5b1e206 ("USB: serial: make minor allocation dynamic") Signed-off-by:
-
Felipe Balbi authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 6c83f772 upstream. If we don't guarantee that we will always get an interrupt at least when we're queueing our very last request, we could fall into situation where we queue every request with 'no_interrupt' set. This will cause the link to get stuck. The behavior above has been triggered with g_ether and dwc3. Reported-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by:
Felipe Balbi <felipe.balbi@linux.intel.com> Signed-off-by:
-
Alexander Usyskin authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 43605e29 upstream. SEC registers are not accessible when the TXE device is in low power state, hence the SEC interrupt cannot be processed if device is not awake. In some rare cases entrance to low power state (aliveness off) and input ready bits can be signaled at the same time, resulting in communication stall as input ready won't be signaled again after waking up. To resolve this IPC_HHIER_SEC bit in HHISR_REG should not be cleaned if the interrupt is not processed. Signed-off-by:
Alexander Usyskin <alexander.usyskin@intel.com> Signed-off-by:
Tomas Winkler <tomas.winkler@intel.com> Signed-off-by:
-
Richard Weinberger authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit a00052a2 upstream. Commit c83ed4c9 ("ubifs: Abort readdir upon error") broke overlayfs support because the fix exposed an internal error code to VFS. Reported-by:
Peter Rosin <peda@axentia.se> Tested-by:
Ralph Sennhauser <ralph.sennhauser@gmail.com> Tested-by:
Richard Weinberger <richard@nod.at> Signed-off-by:
-
Richard Weinberger authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit c83ed4c9 upstream. If UBIFS is facing an error while walking a directory, it reports this error and ubifs_readdir() returns the error code. But the VFS readdir logic does not make the getdents system call fail in all cases. When the readdir cursor indicates that more entries are present, the system call will just return and the libc wrapper will try again since it also knows that more entries are present. This causes the libc wrapper to busy loop for ever when a directory is corrupted on UBIFS. A common approach do deal with corrupted directory entries is skipping them by setting the cursor to the next entry. On UBIFS this approach is not possible since we cannot compute the next directory entry cursor position without reading the current entry. So all we can do is setting the cursor to the "no more entries" position and make getdents exit. Signed-off-by:
-
Chris Mason authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 570dd450 upstream. btrfs_remove_all_log_ctxs takes a shortcut where it avoids walking the list because it knows all of the waiters are patiently waiting for the commit to finish. But, there's a small race where btrfs_sync_log can remove itself from the list if it finds a log commit is already done. Also, it uses list_del_init() to remove itself from the list, but there's no way to know if btrfs_remove_all_log_ctxs has already run, so we don't know for sure if it is safe to call list_del_init(). This gets rid of all the shortcuts for btrfs_remove_all_log_ctxs(), and just calls it with the proper locking. This is part two of the corruption fixed by cbd60aa7. I should have done this in the first place, but convinced myself the optimizations were safe. A 12 hour run of dbench 2048 will eventually trigger a list debug WARN_ON for the list_del_init() in btrfs_sync_log(). Fixes: d1433debReported-by:
Dave Jones <davej@codemonkey.org.uk> Signed-off-by:
Chris Mason <clm@fb.com> Signed-off-by:
-
Arve Hjønnevåg authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 4afb604e upstream. Prevents leaking pointers between processes Signed-off-by:
Arve Hjønnevåg <arve@android.com> Signed-off-by:
Martijn Coenen <maco@android.com> Signed-off-by:
-
Arve Hjønnevåg authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 0a3ffab9 upstream. Prevent using a binder_ref with only weak references where a strong reference is required. Signed-off-by:
-
Hui Wang authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 6aecd871 upstream. They uses the codec ALC255, and have the different pin cfg definition from the ones in the existing pin quirk table. Now adding them into the table to fix the problem. Signed-off-by:
Hui Wang <hui.wang@canonical.com> Signed-off-by:
Takashi Iwai <tiwai@suse.de> Signed-off-by:
-
Hui Wang authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit f771d5bb upstream. We have a new Dell laptop model which uses ALC295, the pin definition is different from the existing ones in the pin quirk table, to fix the headset mic detection and mic mute led's problem, we need to add the new pin defintion into the pin quirk table. Signed-off-by:
-
Ard Biesheuvel authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 3ab7511e upstream. Commit 49d9e77e ("ALSA: hda - Fix system panic when DMA > 40 bits for Nvidia audio controllers") simply disabled any DMA exceeding 32 bits for NVidia devices, even though they are capable of performing DMA up to 40 bits. On some architectures (such as arm64), system memory is not guaranteed to be 32-bit addressable by PCI devices, and so this change prevents NVidia devices from working on platforms such as AMD Seattle. Since the original commit already mentioned that up to 40 bits of DMA is supported, and given that the code has been updated in the meantime to support a 40 bit DMA mask on other devices, revert commit 49d9e77e and explicitly set the DMA mask to 40 bits for NVidia devices. Fixes: 49d9e77e ('ALSA: hda - Fix system panic when DMA > 40 bits...') Signed-off-by:
Ard Biesheuvel <ard.biesheuvel@linaro.org> Signed-off-by:
-
Takashi Iwai authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 7d9a1808 upstream. AZX_DCAPS_RIRB_DELAY is dedicated only for Nvidia and its purpose is just to set a flag in bus. So it's better to be set in the toplevel driver, either hda_intel.c or hda_tegra.c, instead of the common hda_controller.c. This also allows us to strip this flag from dcaps, so save one more bit there. Signed-off-by:
-
Takashi Iwai authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit ef85f299 upstream. AZX_DCAPS_RIRB_PRE_DELAY is always tied with AZX_DCAPS_CTX_WORKAROUND, which is Creative's XFi specific. So, we can replace it and reduce one more bit free for DCAPS. Signed-off-by:
-
Marcel Hasler authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit bdc3478f upstream. The stk1160 chip needs QUIRK_AUDIO_ALIGN_TRANSFER. This patch resolves the issue reported on the mailing list (http://marc.info/?l=linux-sound&m=139223599126215&w=2) and also fixes bug 180071 (https://bugzilla.kernel.org/show_bug.cgi?id=180071). Signed-off-by:
Marcel Hasler <mahasler@gmail.com> Signed-off-by:
-
Johannes Weiner authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 89a28483 upstream. On 4.0, we saw a stack corruption from a page fault entering direct memory cgroup reclaim, calling into btrfs_releasepage(), which then tried to allocate an extent and recursed back into a kmem charge ad nauseam: [...] btrfs_releasepage+0x2c/0x30 try_to_release_page+0x32/0x50 shrink_page_list+0x6da/0x7a0 shrink_inactive_list+0x1e5/0x510 shrink_lruvec+0x605/0x7f0 shrink_zone+0xee/0x320 do_try_to_free_pages+0x174/0x440 try_to_free_mem_cgroup_pages+0xa7/0x130 try_charge+0x17b/0x830 memcg_charge_kmem+0x40/0x80 new_slab+0x2d9/0x5a0 __slab_alloc+0x2fd/0x44f kmem_cache_alloc+0x193/0x1e0 alloc_extent_state+0x21/0xc0 __clear_extent_bit+0x2b5/0x400 try_release_extent_mapping+0x1a3/0x220 __btrfs_releasepage+0x31/0x70 btrfs_releasepage+0x2c/0x30 try_to_release_page+0x32/0x50 shrink_page_list+0x6da/0x7a0 shrink_inactive_list+0x1e5/0x510 shrink_lruvec+0x605/0x7f0 shrink_zone+0xee/0x320 do_try_to_free_pages+0x174/0x440 try_to_free_mem_cgroup_pages+0xa7/0x130 try_charge+0x17b/0x830 mem_cgroup_try_charge+0x65/0x1c0 handle_mm_fault+0x117f/0x1510 __do_page_fault+0x177/0x420 do_page_fault+0xc/0x10 page_fault+0x22/0x30 On later kernels, kmem charging is opt-in rather than opt-out, and that particular kmem allocation in btrfs_releasepage() is no longer being charged and won't recurse and overrun the stack anymore. But it's not impossible for an accounted allocation to happen from the memcg direct reclaim context, and we needed to reproduce this crash many times before we even got a useful stack trace out of it. Like other direct reclaimers, mark tasks in memcg reclaim PF_MEMALLOC to avoid recursing into any other form of direct reclaim. Then let recursive charges from PF_MEMALLOC contexts bypass the cgroup limit. Link: http://lkml.kernel.org/r/20161025141050.GA13019@cmpxchg.orgSigned-off-by:
Johannes Weiner <hannes@cmpxchg.org> Acked-by:
Michal Hocko <mhocko@suse.com> Cc: Vladimir Davydov <vdavydov.dev@gmail.com> Cc: Tejun Heo <tj@kernel.org> Signed-off-by:
Andrew Morton <akpm@linux-foundation.org> Signed-off-by:
Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by:
-
Alexander Polakov authored
BugLink: http://bugs.launchpad.net/bugs/1642572 commit 1bc11d70 upstream. As described in https://bugzilla.kernel.org/show_bug.cgi?id=177821: After some analysis it seems to be that the problem is in alloc_super(). In case list_lru_init_memcg() fails it goes into destroy_super(), which calls list_lru_destroy(). And in list_lru_init() we see that in case memcg_init_list_lru() fails, lru->node is freed, but not set NULL, which then leads list_lru_destroy() to believe it is initialized and call memcg_destroy_list_lru(). memcg_destroy_list_lru() in turn can access lru->node[i].memcg_lrus, which is NULL. [akpm@linux-foundation.org: add comment] Signed-off-by:
Alexander Polakov <apolyakov@beget.ru> Acked-by:
Vladimir Davydov <vdavydov.dev@gmail.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by:
-
