- 27 Sep, 2017 25 commits
-
-
Douglas Leung authored
commit 2cfa5825 upstream. Implement fused multiply-add with correct accuracy. Fused multiply-add operation has better accuracy than respective sequential execution of multiply and add operations applied on the same inputs. This is because accuracy errors accumulate in latter case. This patch implements fused multiply-add with the same accuracy as it is implemented in hardware, using 128-bit intermediate calculations. One test case example (raw bits) that this patch fixes: MADDF.D fd,fs,ft: fd = 0x00000ca000000000 fs = ft = 0x3f40624dd2f1a9fc Fixes: e24c3bec ("MIPS: math-emu: Add support for the MIPS R6 MADDF FPU instruction") Fixes: 83d43305 ("MIPS: math-emu: Add support for the MIPS R6 MSUBF FPU instruction") Signed-off-by:
Douglas Leung <douglas.leung@imgtec.com> Signed-off-by:
Miodrag Dinic <miodrag.dinic@imgtec.com> Signed-off-by:
Goran Ferenc <goran.ferenc@imgtec.com> Signed-off-by:
Aleksandar Markovic <aleksandar.markovic@imgtec.com> Cc: Douglas Leung <douglas.leung@imgtec.com> Cc: Bo Hu <bohu@google.com> Cc: James Hogan <james.hogan@imgtec.com> Cc: Jin Qian <jinqian@google.com> Cc: Paul Burton <paul.burton@imgtec.com> Cc: Petar Jovanovic <petar.jovanovic@imgtec.com> Cc: Raghu Gandham <raghu.gandham@imgtec.com> Cc: linux-mips@linux-mips.org Cc: linux-kernel@vger.kernel.org Patchwork: https://patchwork.linux-mips.org/patch/16891/Signed-off-by:
Ralf Baechle <ralf@linux-mips.org> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Douglas Leung authored
commit b3b8e1eb upstream. Implement fused multiply-add with correct accuracy. Fused multiply-add operation has better accuracy than respective sequential execution of multiply and add operations applied on the same inputs. This is because accuracy errors accumulate in latter case. This patch implements fused multiply-add with the same accuracy as it is implemented in hardware, using 64-bit intermediate calculations. One test case example (raw bits) that this patch fixes: MADDF.S fd,fs,ft: fd = 0x22575225 fs = ft = 0x3727c5ac Fixes: e24c3bec ("MIPS: math-emu: Add support for the MIPS R6 MADDF FPU instruction") Fixes: 83d43305 ("MIPS: math-emu: Add support for the MIPS R6 MSUBF FPU instruction") Signed-off-by:
-
Aleksandar Markovic authored
commit ae11c061 upstream. Fix definition and usage of "maddf_flags" enumeration. Avoid duplicate definition and apply more common capitalization. This patch does not change any scenario. It just makes MADDF and MSUBF emulation code more readable and easier to maintain, and hopefully prevents future bugs as well. Signed-off-by:
James Hogan <james.hogan@imgtec.com> Cc: Bo Hu <bohu@google.com> Cc: Douglas Leung <douglas.leung@imgtec.com> Cc: Jin Qian <jinqian@google.com> Cc: Paul Burton <paul.burton@imgtec.com> Cc: Petar Jovanovic <petar.jovanovic@imgtec.com> Cc: Raghu Gandham <raghu.gandham@imgtec.com> Cc: linux-mips@linux-mips.org Cc: linux-kernel@vger.kernel.org Patchwork: https://patchwork.linux-mips.org/patch/16889/Signed-off-by:
-
Aleksandar Markovic authored
commit 7cf64ce4 upstream. Fix the cases of <MADDF|MSUBF>.<D|S> when any of two multiplicands is +0 or -0, and the third input is also +0 or -0. Depending on the signs of inputs, certain special cases must be handled. A relevant example: MADDF.S fd,fs,ft: If fs contains +0.0, ft contains -0.0, and fd contains 0.0, fd is going to contain +0.0 (without this patch, it used to contain -0.0). Fixes: e24c3bec ("MIPS: math-emu: Add support for the MIPS R6 MADDF FPU instruction") Fixes: 83d43305 ("MIPS: math-emu: Add support for the MIPS R6 MSUBF FPU instruction") Signed-off-by:
-
Aleksandar Markovic authored
commit 0c64fe63 upstream. Fix the cases of <MADDF|MSUBF>.<D|S> when any of two multiplicands is infinity. The correct behavior in such cases is affected by the nature of third input. Cases of addition of infinities with opposite signs and subtraction of infinities with same signs may arise and must be handles separately. Also, the value od flags argument (that determines whether the instruction is MADDF or MSUBF) affects the outcome. Relevant examples: MADDF.S fd,fs,ft: If fs contains +inf, ft contains +inf, and fd contains -inf, fd is going to contain indef (without this patch, it used to contain -inf). MSUBF.S fd,fs,ft: If fs contains +inf, ft contains 1.0, and fd contains +0.0, fd is going to contain -inf (without this patch, it used to contain +inf). Fixes: e24c3bec ("MIPS: math-emu: Add support for the MIPS R6 MADDF FPU instruction") Fixes: 83d43305 ("MIPS: math-emu: Add support for the MIPS R6 MSUBF FPU instruction") Signed-off-by:
-
Aleksandar Markovic authored
commit e840be6e upstream. Fix the cases of <MADDF|MSUBF>.<D|S> when any of three inputs is any NaN. Correct behavior of <MADDF|MSUBF>.<D|S> fd, fs, ft is following: - if any of inputs is sNaN, return a sNaN using following rules: if only one input is sNaN, return that one; if more than one input is sNaN, order of precedence for return value is fd, fs, ft - if no input is sNaN, but at least one of inputs is qNaN, return a qNaN using following rules: if only one input is qNaN, return that one; if more than one input is qNaN, order of precedence for return value is fd, fs, ft The previous code contained correct handling of some above cases, but not all. Also, such handling was scattered into various cases of "switch (CLPAIR(xc, yc))" statement, and elsewhere. With this patch, this logic is placed in one place, and "switch (CLPAIR(xc, yc))" is significantly simplified. A relevant example: MADDF.S fd,fs,ft: If fs contains qNaN1, ft contains qNaN2, and fd contains qNaN3, fd is going to contain qNaN3 (without this patch, it used to contain qNaN1). Fixes: e24c3bec ("MIPS: math-emu: Add support for the MIPS R6 MADDF FPU instruction") Fixes: 83d43305 ("MIPS: math-emu: Add support for the MIPS R6 MSUBF FPU instruction") Signed-off-by:
-
Aleksandar Markovic authored
commit ddbfff74 upstream. If accumulator value is zero, just return the value of previously calculated product. This brings logic in MADDF/MSUBF implementation closer to the logic in ADD/SUB case. Signed-off-by:
-
Aleksandar Markovic authored
commit 304bfe47 upstream. Fix following special cases for MINA>.<D|S>: - if one of the inputs is zero, and the other is subnormal, normal, or infinity, the value of the former should be returned (that is, a zero). - if one of the inputs is infinity, and the other input is normal, or subnormal, the value of the latter should be returned. The previous implementation's logic for such cases was incorrect - it appears as if it implements MAXA, and not MINA instruction. A relevant example: MINA.S fd,fs,ft: If fs contains 100.0, and ft contains 0.0, fd is going to contain 0.0 (without this patch, it used to contain 100.0). Fixes: a79f5f9b ("MIPS: math-emu: Add support for the MIPS R6 MAX{, A} FPU instruction") Fixes: 4e9561b2 ("MIPS: math-emu: Add support for the MIPS R6 MIN{, A} FPU instruction") Signed-off-by:
-
Aleksandar Markovic authored
commit 3444c4eb upstream. Fix the value returned by <MAXA|MINA>.<D|S> fd,fs,ft, if both inputs are infinite. The previous implementation returned always the value contained in ft in such cases. The correct behavior is specified in Mips instruction set manual and is as follows: fs ft MAXA MINA --------------------------------- inf inf inf inf inf -inf inf -inf -inf inf inf -inf -inf -inf -inf -inf A relevant example: MAXA.S fd,fs,ft: If fs contains +inf, and ft contains -inf, fd is going to contain +inf (without this patch, it used to contain -inf). Fixes: a79f5f9b ("MIPS: math-emu: Add support for the MIPS R6 MAX{, A} FPU instruction") Fixes: 4e9561b2 ("MIPS: math-emu: Add support for the MIPS R6 MIN{, A} FPU instruction") Signed-off-by:
-
Aleksandar Markovic authored
commit 1a41b3b4 upstream. Fix the value returned by <MAXA|MINA>.<D|S>, if the inputs are normal fp numbers of the same absolute value, but opposite signs. A relevant example: MAXA.S fd,fs,ft: If fs contains -3.0, and ft contains +3.0, fd is going to contain +3.0 (without this patch, it used to contain -3.0). Fixes: a79f5f9b ("MIPS: math-emu: Add support for the MIPS R6 MAX{, A} FPU instruction") Fixes: 4e9561b2 ("MIPS: math-emu: Add support for the MIPS R6 MIN{, A} FPU instruction") Signed-off-by:
-
Aleksandar Markovic authored
commit aabf5cf0 upstream. Fix the value returned by <MAX|MIN>.<D|S>, if both inputs are negative normal fp numbers. The previous logic did not take into account that if both inputs have the same sign, there should be separate treatment of the cases when both inputs are negative and when both inputs are positive. A relevant example: MAX.S fd,fs,ft: If fs contains -5.0, and ft contains -7.0, fd is going to contain -5.0 (without this patch, it used to contain -7.0). Fixes: a79f5f9b ("MIPS: math-emu: Add support for the MIPS R6 MAX{, A} FPU instruction") Fixes: 4e9561b2 ("MIPS: math-emu: Add support for the MIPS R6 MIN{, A} FPU instruction") Signed-off-by:
-
Aleksandar Markovic authored
commit 15560a58 upstream. Fix the value returned by <MAX|MAXA|MIN|MINA>.<D|S>, if both inputs are zeros. The right behavior in such cases is stated in instruction reference manual and is as follows: fs ft MAX MIN MAXA MINA --------------------------------------------- 0 0 0 0 0 0 0 -0 0 -0 0 -0 -0 0 0 -0 0 -0 -0 -0 -0 -0 -0 -0 Prior to this patch, some of the above cases were yielding correct results. However, for the sake of code consistency, all such cases are rewritten in this patch. A relevant example: MAX.S fd,fs,ft: If fs contains +0.0, and ft contains -0.0, fd is going to contain +0.0 (without this patch, it used to contain -0.0). Fixes: a79f5f9b ("MIPS: math-emu: Add support for the MIPS R6 MAX{, A} FPU instruction") Fixes: 4e9561b2 ("MIPS: math-emu: Add support for the MIPS R6 MIN{, A} FPU instruction") Signed-off-by:
-
Aleksandar Markovic authored
commit e78bf0dc upstream. Fix the value returned by <MAX|MAXA|MIN|MINA>.<D|S> fd,fs,ft, if both inputs are quiet NaNs. The <MAX|MAXA|MIN|MINA>.<D|S> specifications state that the returned value in such cases should be the quiet NaN contained in register fs. A relevant example: MAX.S fd,fs,ft: If fs contains qNaN1, and ft contains qNaN2, fd is going to contain qNaN1 (without this patch, it used to contain qNaN2). Fixes: a79f5f9b ("MIPS: math-emu: Add support for the MIPS R6 MAX{, A} FPU instruction") Fixes: 4e9561b2 ("MIPS: math-emu: Add support for the MIPS R6 MIN{, A} FPU instruction") Signed-off-by:
-
Kai-Heng Feng authored
commit 697c5d8a upstream. Similar to other Gigabyte laptops, the touchpad on P57 requires a keyboard reset to detect Elantech touchpad correctly. BugLink: https://bugs.launchpad.net/bugs/1594214Signed-off-by:
Kai-Heng Feng <kai.heng.feng@canonical.com> Signed-off-by:
Dmitry Torokhov <dmitry.torokhov@gmail.com> Signed-off-by:
-
Daniel Drake authored
commit 79d2c8be upstream. The touchpad in the Asus laptop models X505BA/BP and X542BA/BP is unresponsive after suspend/resume. The following error appears during resume: i2c_hid i2c-ELAN1300:00: failed to reset device. The problem here is that i2c_hid does not notice the interrupt being generated at this point, because the GPIO is no longer configured for interrupts. Fix this by saving pinctrl-amd pin registers during suspend and restoring them at resume time. Based on code from pinctrl-intel. Signed-off-by:
Daniel Drake <drake@endlessm.com> Signed-off-by:
Linus Walleij <linus.walleij@linaro.org> Signed-off-by:
-
Arnd Bergmann authored
commit 8a5a90a2 upstream. Sergey noticed a small but fatal mistake in __tty_insert_flip_char, leading to an oops in an interrupt handler when using any serial port. The problem is that I accidentally took the tty_buffer pointer before calling __tty_buffer_request_room(), which replaces the buffer. This moves the pointer lookup to the right place after allocating the new buffer space. Fixes: 979990c6 ("tty: improve tty_insert_flip_char() fast path") Reported-by:
Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Tested-by:
Arnd Bergmann <arnd@arndb.de> Signed-off-by:
-
Arnd Bergmann authored
commit 065ea0a7 upstream. While working on improving the fast path of tty_insert_flip_char(), I noticed that by calling tty_buffer_request_room(), we needlessly move to the separate flag buffer mode for the tty, even when all characters use TTY_NORMAL as the flag. This changes the code to call __tty_buffer_request_room() with the correct flag, which will then allocate a regular buffer when it rounds out of space but no special flags have been used. I'm guessing that this is the behavior that Peter Hurley intended when he introduced the compacted flip buffers. Fixes: acc0f67f ("tty: Halve flip buffer GFP_ATOMIC memory consumption") Cc: Peter Hurley <peter@hurleysoftware.com> Signed-off-by:
-
Arnd Bergmann authored
commit 979990c6 upstream. kernelci.org reports a crazy stack usage for the VT code when CONFIG_KASAN is enabled: drivers/tty/vt/keyboard.c: In function 'kbd_keycode': drivers/tty/vt/keyboard.c:1452:1: error: the frame size of 2240 bytes is larger than 2048 bytes [-Werror=frame-larger-than=] The problem is that tty_insert_flip_char() gets inlined many times into kbd_keycode(), and also into other functions, and each copy requires 128 bytes for stack redzone to check for a possible out-of-bounds access on the 'ch' and 'flags' arguments that are passed into tty_insert_flip_string_flags as a variable-length string. This introduces a new __tty_insert_flip_char() function for the slow path, which receives the two arguments by value. This completely avoids the problem and the stack usage goes back down to around 100 bytes. Without KASAN, this is also slightly better, as we don't have to spill the arguments to the stack but can simply pass 'ch' and 'flag' in registers, saving a few bytes in .text for each call site. This should be backported to linux-4.0 or later, which first introduced the stack sanitizer in the kernel. Fixes: c420f167 ("kasan: enable stack instrumentation") Signed-off-by:
-
Roland Dreier authored
commit 79e25959 upstream. Commit eea40b8f ("infiniband: call ipv6 route lookup via the stub interface") introduced a regression in address resolution when connecting to IPv6 destination addresses. The old code called ip6_route_output(), while the new code calls ipv6_stub->ipv6_dst_lookup(). The two are almost the same, except that ipv6_dst_lookup() also calls ip6_route_get_saddr() if the source address is in6addr_any. This means that the test of ipv6_addr_any(&fl6.saddr) now never succeeds, and so we never copy the source address out. This ends up causing rdma_resolve_addr() to fail, because without a resolved source address, cma_acquire_dev() will fail to find an RDMA device to use. For me, this causes connecting to an NVMe over Fabrics target via RoCE / IPv6 to fail. Fix this by copying out fl6.saddr if ipv6_addr_any() is true for the original source address passed into addr6_resolve(). We can drop our call to ipv6_dev_get_saddr() because ipv6_dst_lookup() already does that work. Fixes: eea40b8f ("infiniband: call ipv6 route lookup via the stub interface") Signed-off-by:
Roland Dreier <roland@purestorage.com> Acked-by:
Paolo Abeni <pabeni@redhat.com> Signed-off-by:
Doug Ledford <dledford@redhat.com> Signed-off-by:
Raju Rangoju <rajur@chelsio.com> Signed-off-by:
-
Jonathan Liu authored
commit 2a596fc9 upstream. The drm_driver lastclose callback is called when the last userspace DRM client has closed. Call drm_fbdev_cma_restore_mode to restore the fbdev console otherwise the fbdev console will stop working. Fixes: 9026e0d1 ("drm: Add Allwinner A10 Display Engine support") Tested-by:
Olliver Schinagl <oliver@schinagl.nl> Reviewed-by:
Chen-Yu Tsai <wens@csie.org> Signed-off-by:
Jonathan Liu <net147@gmail.com> Signed-off-by:
Maxime Ripard <maxime.ripard@free-electrons.com> [net147@gmail.com: Backport to 4.9, minor context change] Signed-off-by:
-
Mike Marciniszyn authored
commit 5b0ef650 upstream. Section 9.7.7.2.5 of the 1.3 IBTA spec clearly says that receive credits should never apply to RDMA write. qib and hfi1 were doing that. The following situation will result in a QP hang: - A prior SEND or RDMA_WRITE with immmediate consumed the last credit for a QP using RC receive buffer credits - The prior op is acked so there are no more acks - The peer ULP fails to post receive for some reason - An RDMA write sees that the credits are exhausted and waits - The peer ULP posts receive buffers - The ULP posts a send or RDMA write that will be hung The fix is to avoid the credit test for the RDMA write operation. Reviewed-by:
Kaike Wan <kaike.wan@intel.com> Signed-off-by:
Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by:
Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by:
-
Jan Kara authored
commit b5accbb0 upstream. When new directory 'DIR1' is created in a directory 'DIR0' with SGID bit set, DIR1 is expected to have SGID bit set (and owning group equal to the owning group of 'DIR0'). However when 'DIR0' also has some default ACLs that 'DIR1' inherits, setting these ACLs will result in SGID bit on 'DIR1' to get cleared if user is not member of the owning group. Fix the problem by creating __orangefs_set_acl() function that does not call posix_acl_update_mode() and use it when inheriting ACLs. That prevents SGID bit clearing and the mode has been properly set by posix_acl_create() anyway. Fixes: 07393101 CC: stable@vger.kernel.org CC: Mike Marshall <hubcap@omnibond.com> CC: pvfs2-developers@beowulf-underground.org Signed-off-by:
Jan Kara <jack@suse.cz> Signed-off-by:
Mike Marshall <hubcap@omnibond.com> Signed-off-by:
-
Minchan Kim authored
commit 4855e4a7 upstream. There is race between page freeing and unreserved highatomic. CPU 0 CPU 1 free_hot_cold_page mt = get_pfnblock_migratetype set_pcppage_migratetype(page, mt) unreserve_highatomic_pageblock spin_lock_irqsave(&zone->lock) move_freepages_block set_pageblock_migratetype(page) spin_unlock_irqrestore(&zone->lock) free_pcppages_bulk __free_one_page(mt) <- mt is stale By above race, a page on CPU 0 could go non-highorderatomic free list since the pageblock's type is changed. By that, unreserve logic of highorderatomic can decrease reserved count on a same pageblock severak times and then it will make mismatch between nr_reserved_highatomic and the number of reserved pageblock. So, this patch verifies whether the pageblock is highatomic or not and decrease the count only if the pageblock is highatomic. Link: http://lkml.kernel.org/r/1476259429-18279-3-git-send-email-minchan@kernel.orgSigned-off-by:
Minchan Kim <minchan@kernel.org> Acked-by:
Vlastimil Babka <vbabka@suse.cz> Acked-by:
Mel Gorman <mgorman@techsingularity.net> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Sangseok Lee <sangseok.lee@lge.com> Cc: Michal Hocko <mhocko@suse.com> Signed-off-by:
Andrew Morton <akpm@linux-foundation.org> Signed-off-by:
Linus Torvalds <torvalds@linux-foundation.org> Cc: Miles Chen <miles.chen@mediatek.com> Signed-off-by:
-
Trond Myklebust authored
commit ed6473dd upstream. We want to use kthread_stop() in order to ensure the threads are shut down before we tear down the nfs_callback_info in nfs_callback_down. Tested-and-reviewed-by:
Kinglong Mee <kinglongmee@gmail.com> Reported-by:
Trond Myklebust <trond.myklebust@primarydata.com> Signed-off-by:
J. Bruce Fields <bfields@redhat.com> Cc: Jan Hudoba <kernel@jahu.sk> Signed-off-by:
-
Trond Myklebust authored
commit 9e0d8768 upstream. Refactor to separate out the functions of starting and stopping threads so that they can be used in other helpers. Signed-off-by:
-
- 20 Sep, 2017 15 commits
-
-
Greg Kroah-Hartman authored
-
Steffen Klassert authored
commit 36143645 upstream. rt_cookie might be used uninitialized, fix this by initializing it. Fixes: c5cff856 ("ipv6: add rcu grace period before freeing fib6_node") Signed-off-by:
Steffen Klassert <steffen.klassert@secunet.com> Signed-off-by:
David S. Miller <davem@davemloft.net> Cc: Arnd Bergmann <arnd@arndb.de> Signed-off-by:
-
Darrick J. Wong authored
commit 7bf7a193 upstream. Fix up all the compiler warnings that have crept in. Signed-off-by:
Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by:
Christoph Hellwig <hch@lst.de> Cc: Arnd Bergmann <arnd@arndb.de> Signed-off-by:
-
Song Liu authored
commit 9c72a18e upstream. In raid5, there are scenarios where some ios are deferred to a later time, and some IO need a flush to complete. To make sure we make progress with these IOs, we need to call the following functions: flush_deferred_bios(conf); r5l_flush_stripe_to_raid(conf->log); Both of these functions are called in raid5d(), but missing in raid5_do_work(). As a result, these functions are not called when multi-threading (group_thread_cnt > 0) is enabled. This patch adds calls to these function to raid5_do_work(). Note for stable branches: r5l_flush_stripe_to_raid(conf->log) is need for 4.4+ flush_deferred_bios(conf) is only needed for 4.11+ Signed-off-by:
Song Liu <songliubraving@fb.com> Signed-off-by:
Shaohua Li <shli@fb.com> Signed-off-by:
-
Pan Bian authored
commit 6c370590 upstream. In function xfs_test_remount_options(), kfree() is used to free memory allocated by kmem_zalloc(). But it is better to use kmem_free(). Signed-off-by:
Pan Bian <bianpan2016@163.com> Reviewed-by:
-
Christoph Hellwig authored
commit 8353a814 upstream. Our loop in xfs_finish_page_writeback, which iterates over all buffer heads in a page and then calls end_buffer_async_write, which also iterates over all buffers in the page to check if any I/O is in flight is not only inefficient, but also potentially dangerous as end_buffer_async_write can cause the page and all buffers to be freed. Replace it with a single loop that does the work of end_buffer_async_write on a per-page basis. Signed-off-by:
Brian Foster <bfoster@redhat.com> Reviewed-by:
-
Christoph Hellwig authored
commit dd60687e upstream. Reject attempts to set XFLAGS that correspond to di_flags2 inode flags if the inode isn't a v3 inode, because di_flags2 only exists on v3. Signed-off-by:
-
Amir Goldstein authored
commit 47c7d0b1 upstream. When calling into _xfs_log_force{,_lsn}() with a pointer to log_flushed variable, log_flushed will be set to 1 if: 1. xlog_sync() is called to flush the active log buffer AND/OR 2. xlog_wait() is called to wait on a syncing log buffers xfs_file_fsync() checks the value of log_flushed after _xfs_log_force_lsn() call to optimize away an explicit PREFLUSH request to the data block device after writing out all the file's pages to disk. This optimization is incorrect in the following sequence of events: Task A Task B ------------------------------------------------------- xfs_file_fsync() _xfs_log_force_lsn() xlog_sync() [submit PREFLUSH] xfs_file_fsync() file_write_and_wait_range() [submit WRITE X] [endio WRITE X] _xfs_log_force_lsn() xlog_wait() [endio PREFLUSH] The write X is not guarantied to be on persistent storage when PREFLUSH request in completed, because write A was submitted after the PREFLUSH request, but xfs_file_fsync() of task A will be notified of log_flushed=1 and will skip explicit flush. If the system crashes after fsync of task A, write X may not be present on disk after reboot. This bug was discovered and demonstrated using Josef Bacik's dm-log-writes target, which can be used to record block io operations and then replay a subset of these operations onto the target device. The test goes something like this: - Use fsx to execute ops of a file and record ops on log device - Every now and then fsync the file, store md5 of file and mark the location in the log - Then replay log onto device for each mark, mount fs and compare md5 of file to stored value Cc: Christoph Hellwig <hch@lst.de> Cc: Josef Bacik <jbacik@fb.com> Cc: <stable@vger.kernel.org> Signed-off-by:
Amir Goldstein <amir73il@gmail.com> Reviewed-by:
-
Christoph Hellwig authored
commit 742d8429 upstream. Currently flag switching can be used to easily crash the kernel. Disable the per-inode DAX flag until that is sorted out. Signed-off-by:
-
Brian Foster authored
commit 2dd3d709 upstream. The owner change bmbt scan that occurs during extent swap operations does not handle ordered buffer failures. Buffers that cannot be marked ordered must be physically logged so previously dirty ranges of the buffer can be relogged in the transaction. Since the bmbt scan may need to process and potentially log a large number of blocks, we can't expect to complete this operation in a single transaction. Update extent swap to use a permanent transaction with enough log reservation to physically log a buffer. Update the bmbt scan to physically log any buffers that cannot be ordered and to terminate the scan with -EAGAIN. On -EAGAIN, the caller rolls the transaction and restarts the scan. Finally, update the bmbt scan helper function to skip bmbt blocks that already match the expected owner so they are not reprocessed after scan restarts. Signed-off-by:
-
Brian Foster authored
commit a5814bce upstream. Ordered buffers are used in situations where the buffer is not physically logged but must pass through the transaction/logging pipeline for a particular transaction. As a result, ordered buffers are not unpinned and written back until the transaction commits to the log. Ordered buffers have a strict requirement that the target buffer must not be currently dirty and resident in the log pipeline at the time it is marked ordered. If a dirty+ordered buffer is committed, the buffer is reinserted to the AIL but not physically relogged at the LSN of the associated checkpoint. The buffer log item is assigned the LSN of the latest checkpoint and the AIL effectively releases the previously logged buffer content from the active log before the buffer has been written back. If the tail pushes forward and a filesystem crash occurs while in this state, an inconsistent filesystem could result. It is currently the caller responsibility to ensure an ordered buffer is not already dirty from a previous modification. This is unclear and error prone when not used in situations where it is guaranteed a buffer has not been previously modified (such as new metadata allocations). To facilitate general purpose use of ordered buffers, update xfs_trans_ordered_buf() to conditionally order the buffer based on state of the log item and return the status of the result. If the bli is dirty, do not order the buffer and return false. The caller must either physically log the buffer (having acquired the appropriate log reservation) or push it from the AIL to clean it before it can be marked ordered in the current transaction. Note that ordered buffers are currently only used in two situations: 1.) inode chunk allocation where previously logged buffers are not possible and 2.) extent swap which will be updated to handle ordered buffer failures in a separate patch. Signed-off-by:
-
Brian Foster authored
commit 6fb10d6d upstream. The extent swap operation currently resets bmbt block owners before the inode forks are swapped. The bmbt buffers are marked as ordered so they do not have to be physically logged in the transaction. This use of ordered buffers is not safe as bmbt buffers may have been previously physically logged. The bmbt owner change algorithm needs to be updated to physically log buffers that are already dirty when/if they are encountered. This means that an extent swap will eventually require multiple rolling transactions to handle large btrees. In addition, all inode related changes must be logged before the bmbt owner change scan begins and can roll the transaction for the first time to preserve fs consistency via log recovery. In preparation for such fixes to the bmbt owner change algorithm, refactor the bmbt scan out of the extent fork swap code to the last operation before the transaction is committed. Update xfs_swap_extent_forks() to only set the inode log flags when an owner change scan is necessary. Update xfs_swap_extents() to trigger the owner change based on the inode log flags. Note that since the owner change now occurs after the extent fork swap, the inode btrees must be fixed up with the inode number of the current inode (similar to log recovery). Signed-off-by:
-
Brian Foster authored
commit 99c794c6 upstream. Extent swap uses xfs_btree_visit_blocks() to fix up bmbt block owners on v5 (!rmapbt) filesystems. The bmbt scan uses xfs_btree_lookup_get_block() to read bmbt blocks which verifies the current owner of the block against the parent inode of the bmbt. This works during extent swap because the bmbt owners are updated to the opposite inode number before the inode extent forks are swapped. The modified bmbt blocks are marked as ordered buffers which allows everything to commit in a single transaction. If the transaction commits to the log and the system crashes such that recovery of the extent swap is required, log recovery restarts the bmbt scan to fix up any bmbt blocks that may have not been written back before the crash. The log recovery bmbt scan occurs after the inode forks have been swapped, however. This causes the bmbt block owner verification to fail, leads to log recovery failure and requires xfs_repair to zap the log to recover. Define a new invalid inode owner flag to inform the btree block lookup mechanism that the current inode may be invalid with respect to the current owner of the bmbt block. Set this flag on the cursor used for change owner scans to allow this operation to work at runtime and during log recovery. Signed-off-by:
-
Brian Foster authored
commit 8dc518df upstream. Ordered buffers are attached to transactions and pushed through the logging infrastructure just like normal buffers with the exception that they are not actually written to the log. Therefore, we don't need to log dirty ranges of ordered buffers. xfs_trans_log_buf() is called on ordered buffers to set up all of the dirty state on the transaction, buffer and log item and prepare the buffer for I/O. Now that xfs_trans_dirty_buf() is available, call it from xfs_trans_ordered_buf() so the latter is now mutually exclusive with xfs_trans_log_buf(). This reflects the implementation of ordered buffers and helps eliminate confusion over the need to log ranges of ordered buffers just to set up internal log state. Signed-off-by:
Allison Henderson <allison.henderson@oracle.com> Reviewed-by:
-
Brian Foster authored
commit 9684010d upstream. xfs_trans_log_buf() is responsible for logging the dirty segments of a buffer along with setting all of the necessary state on the transaction, buffer, bli, etc., to ensure that the associated items are marked as dirty and prepared for I/O. We have a couple use cases that need to to dirty a buffer in a transaction without actually logging dirty ranges of the buffer. One existing use case is ordered buffers, which are currently logged with arbitrary ranges to accomplish this even though the content of ordered buffers is never written to the log. Another pending use case is to relog an already dirty buffer across rolled transactions within the deferred operations infrastructure. This is required to prevent a held (XFS_BLI_HOLD) buffer from pinning the tail of the log. Refactor xfs_trans_log_buf() into a new function that contains all of the logic responsible to dirty the transaction, lidp, buffer and bli. This new function can be used in the future for the use cases outlined above. This patch does not introduce functional changes. Signed-off-by:
-
