- 19 Jan, 2017 17 commits
-
-
Laurent Pinchart authored
The field isn't set by any platform but is only used internally in the driver to hold data parsed from DT. Move it to the sci_port structure. Signed-off-by:
Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com> Reviewed-by:
Geert Uytterhoeven <geert+renesas@glider.be> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Laurent Pinchart authored
The sh-sci driver implements manual break debouncing for a few SH platforms by reading the value of the RX pin port register. This feature is optional and the driver considers all negative or zero values of the platform data port_reg field as invalid. As the four platforms that set the field to a register address all use an address higher than 0x7fffffff, the driver will always consider the value as invalid and never perform debouncing. The feature is unused, remove it. Debouncing could be implemented properly in the future using the pinctrl and GPIO APIs if desired. Signed-off-by:
-
Laurent Pinchart authored
Only SH platforms still use platform data for the sh-sci, and none of them declare DMA channels connected to the SCI. Remove the corresponding platform data fields and simplify the driver accordingly. Signed-off-by:
-
Laurent Pinchart authored
The fifo size, overrun register and mask, sampling rate mask and error mask all depend on the port type only and don't need to be computed at runtime. Add them to the sci_port_parameters structure. Signed-off-by:
-
Laurent Pinchart authored
The driver modifies platform data for internal purpose only. Fix that and make the platform data structure const. Signed-off-by:
-
Laurent Pinchart authored
Turn the regmap two-dimensional array to an array of port parameters and store a pointer to the port parameters in the sci_port structure. This will allow handling additional port type dependent parameters. Signed-off-by:
-
Laurent Pinchart authored
The compiler zeros uninitialized fields, don't zero them manually. Signed-off-by:
-
Laurent Pinchart authored
Even though most of its registers are 8-bit wide, the IRDA has two 16-bit registers that make it a 16-bit peripheral and not a 8-bit peripheral with addresses shifted by one. Fix the registers offset in the driver and the platform data regshift value. Signed-off-by:
-
Laurent Pinchart authored
According to the datasheets, the sh7760 SIM and sh7723 SCIFA instances don't implement the REIE bit. Don't set it in platform data. Signed-off-by:
-
Laurent Pinchart authored
The bit is only set by platforms that also set the CKE1 but, in which case its value is ignored by the device. Don't set it, this simplifies platform data and only leaves the CKE1 bit to be handled. Signed-off-by:
-
Laurent Pinchart authored
The driver considers all negative or zero values of the port_reg field as invalid. The four platforms that set the field to a register address all use an address higher than 0x7fffffff, which is thus considered by the driver as invalid. The feature is thus never used, remove it. The feature could be implemented properly in the future using the pinctrl and GPIO APIs if desired. While at it, don't set the field to SCIx_NOT_SUPPORTED (-1) either, leaving it unset leads to the same result. This will allow removing the SCIx_NOT_SUPPORTED macro. Signed-off-by:
-
Laurent Pinchart authored
The flag is set by the driver internally, don't set it in platform data. Signed-off-by:
-
Laurent Pinchart authored
The SCIF ports on sh7264 and sh7269 don't support the TOIE bit according to the datasheets. Signed-off-by:
-
Laurent Pinchart authored
The scscr platform data field is used by the driver in three locations. One of them masks out all bits except SCSCR_REIE. The two other are the set_termios handler and the console write handler. The set_termios handler calls sci_start_rx() to enable the receiver, which sets the RIE bit unconditionally. It then calls sci_port_disable() that effectively disables both the transmitter and the receiver. The TIE bit will thus get set later when the serial cores reenables the serial port. The console write handler runs with interrupts disabled, and saves and restores the SCSCR register value. The RIE and TIE bits are thus not needed there. The bits are thus not necessary in platform data, remove them. Signed-off-by:
-
Laurent Pinchart authored
The bits are set by the driver internally, don't set them in platform data. Signed-off-by:
-
Laurent Pinchart authored
The UPF_BOOT_AUTOCONF platform data flag is set by all platforms, hardcode it. The UPF_IOREMAP flag is set by a single SH platform and thus needs to be kept. However, for ARM platforms, we can base the decision on whether an OF node is present and bypass the platform data flags completely. Signed-off-by:
-
Laurent Pinchart authored
The Transmit Enable and Receive Enable bits are set in the scscr field of all instances of the sh-sci platform data. Set them in the driver directly to prepare for their removal from platform data. Signed-off-by:
-
- 16 Jan, 2017 3 commits
-
-
Greg Kroah-Hartman authored
We want the serial/tty fixes in here as well to build on top of. Signed-off-by: -
Linus Torvalds authored
-
git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespaceLinus Torvalds authored
Pull namespace fixes from Eric Biederman: "This tree contains 4 fixes. The first is a fix for a race that can causes oopses under the right circumstances, and that someone just recently encountered. Past that are several small trivial correct fixes. A real issue that was blocking development of an out of tree driver, but does not appear to have caused any actual problems for in-tree code. A potential deadlock that was reported by lockdep. And a deadlock people have experienced and took the time to track down caused by a cleanup that removed the code to drop a reference count" * 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: sysctl: Drop reference added by grab_header in proc_sys_readdir pid: fix lockdep deadlock warning due to ucount_lock libfs: Modify mount_pseudo_xattr to be clear it is not a userspace mount mnt: Protect the mountpoint hashtable with mount_lock
-
- 15 Jan, 2017 14 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-miscLinus Torvalds authored
Pull char/misc driver fixes from Greg KH: "Here are some small char/misc driver fixes for 4.10-rc4 that resolve some reported issues. The MEI driver issue resolves a lot of problems that people have been having, as does the mem driver fix. The other minor fixes resolve other reported issues. All of these have been in linux-next for a while" * tag 'char-misc-4.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc: vme: Fix wrong pointer utilization in ca91cx42_slave_get auxdisplay: fix new ht16k33 build errors ppdev: don't print a free'd string extcon: return error code on failure drivers: char: mem: Fix thinkos in kmem address checks mei: bus: enable OS version only for SPT and newer
-
Linus Torvalds authored
Merge tag 'driver-core-4.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core Pull driver core fix from Greg KH: "Here is a single patch being reverted to remove a feature that was added in 4.10-rc1 that isn't quite ready for release. It will be redone as a debugfs file instead of a sysfs file in the future" * tag 'driver-core-4.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core: Revert "driver core: Add deferred_probe attribute to devices in sysfs"
-
git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/ttyLinus Torvalds authored
Pull tty/serial fixes from Greg KH: "Here are some small tty/serial driver fixes for 4.10-rc4 to resolve a number of reported issues. Nothing major here at all, one revert of a problematic patch, and some other tiny bugfixes. Full details are in the shortlog below. All have been in linux-next with no reported issues" * tag 'tty-4.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty: sysrq: attach sysrq handler correctly for 32-bit kernel Revert "tty: serial: 8250: add CON_CONSDEV to flags" Clearing FIFOs in RS485 emulation mode causes subsequent transmits to break 8250_pci: Fix potential use-after-free in error path tty/serial: atmel: RS485 half duplex w/DMA: enable RX after TX is done tty/serial: atmel_serial: BUG: stop DMA from transmitting in stop_tx
-
git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usbLinus Torvalds authored
Pull USB fixes from Greg KH: "Here are a few small USB driver fixes for 4.10-rc4 to resolve some reported issues. The "largest" here is a number of bugs being fixed in the ch341 usb-serial driver, to hopefully resolve the mess of different devices floating around that use this driver that have been having problems with the 4.10-rc1 release. There's also a tiny musb fix that I missed in the last pull request, as well as the traditional xhci fix rounding out the batch. All have been in linux-next with no reported issues" * tag 'usb-4.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: xhci: fix deadlock at host remove by running watchdog correctly USB: serial: ch341: fix control-message error handling usb: musb: fix runtime PM in debugfs wusbcore: Fix one more crypto-on-the-stack bug USB: serial: kl5kusb105: fix line-state error handling USB: serial: ch341: fix baud rate and line-control handling USB: serial: ch341: fix line settings after reset-resume USB: serial: ch341: fix resume after reset USB: serial: ch341: fix open error handling USB: serial: ch341: fix modem-control and B0 handling USB: serial: ch341: fix open and resume after B0 USB: serial: ch341: fix initial modem-control state
-
git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linuxLinus Torvalds authored
Pull i2c fixes from Wolfram Sang: "Bugfixes for I2C. Mostly core this time which is a bit unusual but nothing really scary in there" * 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux: i2c: piix4: Avoid race conditions with IMC i2c: fix spelling mistake: "insufficent" -> "insufficient" i2c: print correct device invalid address i2c: do not enable fall back to Host Notify by default i2c: fix kernel memory disclosure in dev interface
-
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tipLinus Torvalds authored
Pull x86 fixes from Ingo Molnar: "Misc fixes: - unwinder fixes - AMD CPU topology enumeration fixes - microcode loader fixes - x86 embedded platform fixes - fix for a bootup crash that may trigger when clearcpuid= is used with invalid values" * 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: x86/mpx: Use compatible types in comparison to fix sparse error x86/tsc: Add the Intel Denverton Processor to native_calibrate_tsc() x86/entry: Fix the end of the stack for newly forked tasks x86/unwind: Include __schedule() in stack traces x86/unwind: Disable KASAN checks for non-current tasks x86/unwind: Silence warnings for non-current tasks x86/microcode/intel: Use correct buffer size for saving microcode data x86/microcode/intel: Fix allocation size of struct ucode_patch x86/microcode/intel: Add a helper which gives the microcode revision x86/microcode: Use native CPUID to tickle out microcode revision x86/CPU: Add native CPUID variants returning a single datum x86/boot: Add missing declaration of string functions x86/CPU/AMD: Fix Bulldozer topology x86/platform/intel-mid: Rename 'spidev' to 'mrfld_spidev' x86/cpu: Fix typo in the comment for Anniedale x86/cpu: Fix bootup crashes by sanitizing the argument of the 'clearcpuid=' command-line option -
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tipLinus Torvalds authored
Pull NOHZ fix from Ingo Molnar: "This fixes an old NOHZ race where we incorrectly calculate the next timer interrupt in certain circumstances where hrtimers are pending, that can cause hard to reproduce stalled-values artifacts in /proc/stat" * 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: nohz: Fix collision between tick and other hrtimers
-
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tipLinus Torvalds authored
Pull perf fixes from Ingo Molnar: "Misc race fixes uncovered by fuzzing efforts, a Sparse fix, two PMU driver fixes, plus miscellanous tooling fixes" * 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: perf/x86: Reject non sampling events with precise_ip perf/x86/intel: Account interrupts for PEBS errors perf/core: Fix concurrent sys_perf_event_open() vs. 'move_group' race perf/core: Fix sys_perf_event_open() vs. hotplug perf/x86/intel: Use ULL constant to prevent undefined shift behaviour perf/x86/intel/uncore: Fix hardcoded socket 0 assumption in the Haswell init code perf/x86: Set pmu->module in Intel PMU modules perf probe: Fix to probe on gcc generated symbols for offline kernel perf probe: Fix --funcs to show correct symbols for offline module perf symbols: Robustify reading of build-id from sysfs perf tools: Install tools/lib/traceevent plugins with install-bin tools lib traceevent: Fix prev/next_prio for deadline tasks perf record: Fix --switch-output documentation and comment perf record: Make __record_options static tools lib subcmd: Add OPT_STRING_OPTARG_SET option perf probe: Fix to get correct modname from elf header samples/bpf trace_output_user: Remove duplicate sys/ioctl.h include samples/bpf sock_example: Avoid getting ethhdr from two includes perf sched timehist: Show total scheduling time
-
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tipLinus Torvalds authored
Pull EFI fixes from Ingo Molnar: "A number of regression fixes: - Fix a boot hang on machines that have somewhat unusual memory map entries of phys_addr=0x0 num_pages=0, which broke due to a recent commit. This commit got cherry-picked from the v4.11 queue because the bug is affecting real machines. - Fix a boot hang also reported by KASAN, caused by incorrect init ordering introduced by a recent optimization. - Fix a recent robustification fix to allocate_new_fdt_and_exit_boot() that introduced an invalid assumption. Neither bugs were seen in the wild AFAIK" * 'efi-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: efi/x86: Prune invalid memory map entries and fix boot regression x86/efi: Don't allocate memmap through memblock after mm_init() efi/libstub/arm*: Pass latest memory map to the kernel -
git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfsLinus Torvalds authored
Pull vfs fixes from Al Viro. The most notable fix here is probably the fix for a splice regression ("fix a fencepost error in pipe_advance()") noticed by Alan Wylie. * 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: fix a fencepost error in pipe_advance() coredump: Ensure proper size of sparse core files aio: fix lock dep warning tmpfs: clear S_ISGID when setting posix ACLs -
git://git.kernel.dk/linux-blockLinus Torvalds authored
Pull block fixes from Jens Axboe: - the virtio_blk stack DMA corruption fix from Christoph, fixing and issue with VMAP stacks. - O_DIRECT blkbits calculation fix from Chandan. - discard regression fix from Christoph. - queue init error handling fixes for nbd and virtio_blk, from Omar and Jeff. - two small nvme fixes, from Christoph and Guilherme. - rename of blk_queue_zone_size and bdev_zone_size to _sectors instead, to more closely follow what we do in other places in the block layer. This interface is new for this series, so let's get the naming right before releasing a kernel with this feature. From Damien. * 'for-linus' of git://git.kernel.dk/linux-block: block: don't try to discard from __blkdev_issue_zeroout sd: remove __data_len hack for WRITE SAME nvme: use blk_rq_payload_bytes scsi: use blk_rq_payload_bytes block: add blk_rq_payload_bytes block: Rename blk_queue_zone_size and bdev_zone_size nvme: apply DELAY_BEFORE_CHK_RDY quirk at probe time too nvme-rdma: fix nvme_rdma_queue_is_ready virtio_blk: fix panic in initialization error path nbd: blk_mq_init_queue returns an error code on failure, not NULL virtio_blk: avoid DMA to stack for the sense buffer do_direct_IO: Use inode->i_blkbits to compute block count to be cleaned
-
Al Viro authored
The logics in pipe_advance() used to release all buffers past the new position failed in cases when the number of buffers to release was equal to pipe->buffers. If that happened, none of them had been released, leaving pipe full. Worse, it was trivial to trigger and we end up with pipe full of uninitialized pages. IOW, it's an infoleak. Cc: stable@vger.kernel.org # v4.9 Reported-by:
"Alan J. Wylie" <alan@wylie.me.uk> Tested-by:
Al Viro <viro@zeniv.linux.org.uk>
-
Dave Kleikamp authored
If the last section of a core file ends with an unmapped or zero page, the size of the file does not correspond with the last dump_skip() call. gdb complains that the file is truncated and can be confusing to users. After all of the vma sections are written, make sure that the file size is no smaller than the current file position. This problem can be demonstrated with gdb's bigcore testcase on the sparc architecture. Signed-off-by:
Dave Kleikamp <dave.kleikamp@oracle.com> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: linux-fsdevel@vger.kernel.org Cc: linux-kernel@vger.kernel.org Signed-off-by:
-
Shaohua Li authored
lockdep reports a warnning. file_start_write/file_end_write only acquire/release the lock for regular files. So checking the files in aio side too. [ 453.532141] ------------[ cut here ]------------ [ 453.533011] WARNING: CPU: 1 PID: 1298 at ../kernel/locking/lockdep.c:3514 lock_release+0x434/0x670 [ 453.533011] DEBUG_LOCKS_WARN_ON(depth <= 0) [ 453.533011] Modules linked in: [ 453.533011] CPU: 1 PID: 1298 Comm: fio Not tainted 4.9.0+ #964 [ 453.533011] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.9.0-1.fc24 04/01/2014 [ 453.533011] ffff8803a24b7a70 ffffffff8196cffb ffff8803a24b7ae8 0000000000000000 [ 453.533011] ffff8803a24b7ab8 ffffffff81091ee1 ffff8803a5dba700 00000dba00000008 [ 453.533011] ffffed0074496f59 ffff8803a5dbaf54 ffff8803ae0f8488 fffffffffffffdef [ 453.533011] Call Trace: [ 453.533011] [<ffffffff8196cffb>] dump_stack+0x67/0x9c [ 453.533011] [<ffffffff81091ee1>] __warn+0x111/0x130 [ 453.533011] [<ffffffff81091f97>] warn_slowpath_fmt+0x97/0xb0 [ 453.533011] [<ffffffff81091f00>] ? __warn+0x130/0x130 [ 453.533011] [<ffffffff8191b789>] ? blk_finish_plug+0x29/0x60 [ 453.533011] [<ffffffff811205d4>] lock_release+0x434/0x670 [ 453.533011] [<ffffffff8198af94>] ? import_single_range+0xd4/0x110 [ 453.533011] [<ffffffff81322195>] ? rw_verify_area+0x65/0x140 [ 453.533011] [<ffffffff813aa696>] ? aio_write+0x1f6/0x280 [ 453.533011] [<ffffffff813aa6c9>] aio_write+0x229/0x280 [ 453.533011] [<ffffffff813aa4a0>] ? aio_complete+0x640/0x640 [ 453.533011] [<ffffffff8111df20>] ? debug_check_no_locks_freed+0x1a0/0x1a0 [ 453.533011] [<ffffffff8114793a>] ? debug_lockdep_rcu_enabled.part.2+0x1a/0x30 [ 453.533011] [<ffffffff81147985>] ? debug_lockdep_rcu_enabled+0x35/0x40 [ 453.533011] [<ffffffff812a92be>] ? __might_fault+0x7e/0xf0 [ 453.533011] [<ffffffff813ac9bc>] do_io_submit+0x94c/0xb10 [ 453.533011] [<ffffffff813ac2ae>] ? do_io_submit+0x23e/0xb10 [ 453.533011] [<ffffffff813ac070>] ? SyS_io_destroy+0x270/0x270 [ 453.533011] [<ffffffff8111d7b3>] ? mark_held_locks+0x23/0xc0 [ 453.533011] [<ffffffff8100201a>] ? trace_hardirqs_on_thunk+0x1a/0x1c [ 453.533011] [<ffffffff813acb90>] SyS_io_submit+0x10/0x20 [ 453.533011] [<ffffffff824f96aa>] entry_SYSCALL_64_fastpath+0x18/0xad [ 453.533011] [<ffffffff81119190>] ? trace_hardirqs_off_caller+0xc0/0x110 [ 453.533011] ---[ end trace b2fbe664d1cc0082 ]--- Cc: Dmitry Monakhov <dmonakhov@openvz.org> Cc: Jan Kara <jack@suse.cz> Cc: Christoph Hellwig <hch@lst.de> Cc: Al Viro <viro@zeniv.linux.org.uk> Reviewed-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
Shaohua Li <shli@fb.com> Signed-off-by:
-
- 14 Jan, 2017 6 commits
-
-
git://git.infradead.org/users/vkoul/slave-dmaLinus Torvalds authored
Pull dmaengine fixes from Vinod Koul: "The fixes this time around are spread over drivers, pretty normal update: - PCI ID for SKL ioatdma, workaround for SKX and ioat_alloc_chan_resources sleepy allocation fix - dw kconfig typo fix - null pointer deref for stm32 - MAINTAINERS Update for at_hdmac - pl330 runtime pm fixes - omap-dma port window fix - rcar-dmac unmap slave resource fix" * tag 'dmaengine-fix-4.10-rc4' of git://git.infradead.org/users/vkoul/slave-dma: dmaengine: rcar-dmac: unmap slave resource when channel is freed dmaengine: omap-dma: Fix the port_window support dmaengine: iota: ioat_alloc_chan_resources should not perform sleeping allocations. dmaengine: pl330: Fix runtime PM support for terminated transfers MAINTAINERS: dmaengine: Update + Hand over the at_hdmac driver to Ludovic dmaengine: omap-dma: Fix dynamic lch_map allocation dmaengine: ti-dma-crossbar: Add some 'of_node_put()' in error path. dmaengine: stm32-dma: Fix null pointer dereference in stm32_dma_tx_status dmaengine: stm32-dma: Set correct args number for DMA request from DT dmaengine: dw: fix typo in Kconfig dmaengine: ioatdma: workaround SKX ioatdma version dmaengine: ioatdma: Add Skylake PCI Dev ID -
Peter Jones authored
Some machines, such as the Lenovo ThinkPad W541 with firmware GNET80WW (2.28), include memory map entries with phys_addr=0x0 and num_pages=0. These machines fail to boot after the following commit, commit 8e80632f ("efi/esrt: Use efi_mem_reserve() and avoid a kmalloc()") Fix this by removing such bogus entries from the memory map. Furthermore, currently the log output for this case (with efi=debug) looks like: [ 0.000000] efi: mem45: [Reserved | | | | | | | | | | | | ] range=[0x0000000000000000-0xffffffffffffffff] (0MB) This is clearly wrong, and also not as informative as it could be. This patch changes it so that if we find obviously invalid memory map entries, we print an error and skip those entries. It also detects the display of the address range calculation overflow, so the new output is: [ 0.000000] efi: [Firmware Bug]: Invalid EFI memory map entries: [ 0.000000] efi: mem45: [Reserved | | | | | | | | | | | | ] range=[0x0000000000000000-0x0000000000000000] (invalid) It also detects memory map sizes that would overflow the physical address, for example phys_addr=0xfffffffffffff000 and num_pages=0x0200000000000001, and prints: [ 0.000000] efi: [Firmware Bug]: Invalid EFI memory map entries: [ 0.000000] efi: mem45: [Reserved | | | | | | | | | | | | ] range=[phys_addr=0xfffffffffffff000-0x20ffffffffffffffff] (invalid) It then removes these entries from the memory map. Signed-off-by:
Peter Jones <pjones@redhat.com> Signed-off-by:
Ard Biesheuvel <ard.biesheuvel@linaro.org> [ardb: refactor for clarity with no functional changes, avoid PAGE_SHIFT] Signed-off-by:
Matt Fleming <matt@codeblueprint.co.uk> [Matt: Include bugzilla info in commit log] Cc: <stable@vger.kernel.org> # v4.9+ Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: https://bugzilla.kernel.org/show_bug.cgi?id=191121Signed-off-by:
Ingo Molnar <mingo@kernel.org>
-
Greg Kroah-Hartman authored
This reverts commit 6751667a. Rob Herring objected to it, and a replacement for it will be added using debugfs in the future. Cc: Ben Hutchings <ben.hutchings@codethink.co.uk> Reported-by:
Rob Herring <robh@kernel.org> Signed-off-by:
-
Jiri Olsa authored
As Peter suggested [1] rejecting non sampling PEBS events, because they dont make any sense and could cause bugs in the NMI handler [2]. [1] http://lkml.kernel.org/r/20170103094059.GC3093@worktop [2] http://lkml.kernel.org/r/1482931866-6018-3-git-send-email-jolsa@kernel.orgSigned-off-by:
Jiri Olsa <jolsa@redhat.com> Signed-off-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vince Weaver <vince@deater.net> Cc: Vince Weaver <vincent.weaver@maine.edu> Link: http://lkml.kernel.org/r/20170103142454.GA26251@kravaSigned-off-by:
-
Jiri Olsa authored
It's possible to set up PEBS events to get only errors and not any data, like on SNB-X (model 45) and IVB-EP (model 62) via 2 perf commands running simultaneously: taskset -c 1 ./perf record -c 4 -e branches:pp -j any -C 10 This leads to a soft lock up, because the error path of the intel_pmu_drain_pebs_nhm() does not account event->hw.interrupt for error PEBS interrupts, so in case you're getting ONLY errors you don't have a way to stop the event when it's over the max_samples_per_tick limit: NMI watchdog: BUG: soft lockup - CPU#22 stuck for 22s! [perf_fuzzer:5816] ... RIP: 0010:[<ffffffff81159232>] [<ffffffff81159232>] smp_call_function_single+0xe2/0x140 ... Call Trace: ? trace_hardirqs_on_caller+0xf5/0x1b0 ? perf_cgroup_attach+0x70/0x70 perf_install_in_context+0x199/0x1b0 ? ctx_resched+0x90/0x90 SYSC_perf_event_open+0x641/0xf90 SyS_perf_event_open+0x9/0x10 do_syscall_64+0x6c/0x1f0 entry_SYSCALL64_slow_path+0x25/0x25 Add perf_event_account_interrupt() which does the interrupt and frequency checks and call it from intel_pmu_drain_pebs_nhm()'s error path. We keep the pending_kill and pending_wakeup logic only in the __perf_event_overflow() path, because they make sense only if there's any data to deliver. Signed-off-by:Jiri Olsa <jolsa@kernel.org> Signed-off-by:
-
Peter Zijlstra authored
Di Shen reported a race between two concurrent sys_perf_event_open() calls where both try and move the same pre-existing software group into a hardware context. The problem is exactly that described in commit: f63a8daa ("perf: Fix event->ctx locking") ... where, while we wait for a ctx->mutex acquisition, the event->ctx relation can have changed under us. That very same commit failed to recognise sys_perf_event_context() as an external access vector to the events and thereby didn't apply the established locking rules correctly. So while one sys_perf_event_open() call is stuck waiting on mutex_lock_double(), the other (which owns said locks) moves the group about. So by the time the former sys_perf_event_open() acquires the locks, the context we've acquired is stale (and possibly dead). Apply the established locking rules as per perf_event_ctx_lock_nested() to the mutex_lock_double() for the 'move_group' case. This obviously means we need to validate state after we acquire the locks. Reported-by: Di Shen (Keen Lab) Tested-by:
John Dias <joaodias@google.com> Signed-off-by:
-
