On Wed, Jul 08, 2015 at 11:54:04PM -0700, Ming Lin wrote: > On Wed, 2015-07-08 at 16:40 -0700, Ming Lin wrote: > > On Wed, 2015-07-08 at 06:57 -0700, Ming Lin wrote: > > > On Wed, 2015-07-08 at 10:53 +0200, Christoph Hellwig wrote: > > > > On Wed, Jul 08, 2015 at 12:01:53AM -0700, Ming Lin wrote: > > > > > Hi list, > > > > > > > > > > I am testing some direct-io patches. > > > > > xfs_tests/xfs/080 failed. > > > > > > > > > > It's strange. > > > > > There is no hole in "rwtest.file", > > > > > but when read it at some offset it returns zero because of hole. > > > > > > > > > > root@block:~# xfs_io -c bmap /mnt/test/rwtest.file > > > > > /mnt/test/rwtest.file: > > > > > 0: [0..125671]: 160..125831 > > > > > > > > Sounds like you see an unwritten extent. What does xfs_bmap -pv show on > > > > the file? > > > > > > root@block:~# xfs_bmap -pv /mnt/test/rwtest.file > > > /mnt/test/rwtest.file: > > > EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL FLAGS > > > 0: [0..743]: 160..903 0 (160..903) 744 00000 > > > 1: [744..783]: 904..943 0 (904..943) 40 10000 > > > 2: [784..951]: 944..1111 0 (944..1111) 168 00000 > > > 3: [952..1023]: 1112..1183 0 (1112..1183) 72 10000 > > > 4: [1024..1287]: 1184..1447 0 (1184..1447) 264 00000 > > > 5: [1288..1383]: 1448..1543 0 (1448..1543) 96 10000 > > > > file offset 659456 is in EXT 5. > > I guess FLAGS 1 means it's an unwritten extent, right? > > > > xfs_tests/xfs/080 basically does: > > 1. direct io write > > 2. then buffered read to check the previous dio write > > > > Here is an example of error logs. > > > > doio ( 3626) 16:13:23 > > --------------------- > > *** DATA COMPARISON ERROR *** > > check_file(/mnt/test/rwtest.file, 239616, 81408, H:3626:block:doio*, > > 18, 0) failed > > > > Comparison fd is 5, with open flags 0 > > Corrupt regions follow - unprintable chars are represented as '.' > > ----------------------------------------------------------------- > > corrupt bytes starting at file offset 241664 > > 1st 32 expected bytes: oio*H:3626:block:doio*H:3626:blo > > 1st 32 actual bytes: ................................ > > > > Request number 981 > > fd 4 is file /mnt/test/rwtest.file - open flags are 040001 > > O_WRONLY,O_DIRECT, > > write done at file offset 239616 - pattern is H (0110) > > number of requests is 1, strides per request is 1 > > i/o byte count = 81408 > > memory alignment is aligned > > DIRECT I/O: offset % 512 = 0 length % 81408 = 0 > > mem alignment 0x200 xfer size: small: 512 large: 2147483136 > > > > syscall: writev(4, (iov on stack), 1) > > > > > > The "expected bytes" was actually written to the disk successfully, > > but for some bug the xfs extent flag was not set correctly. > > The flag should be "XFS_EXT_NORM", but now it's "XFS_EXT_UNWRITTEN". > > > > Could you share the code path the extent flag was updated? > > Is it also updated through direct io? > > For XFS direct IO, are below the only 2 paths to convert unwritten > extent to written extent? > > [ 72.279043] Call Trace: > [ 72.279719] [<ffffffff8169c2f3>] dump_stack+0x4f/0x7b > [ 72.280898] [<ffffffff8127140f>] xfs_bmapi_convert_unwritten+0x73/0x176 > [ 72.282393] [<ffffffff8127a233>] xfs_bmapi_write+0x2a0/0x7ee > [ 72.283705] [<ffffffff812b8ae8>] xfs_iomap_write_direct+0x23e/0x2f0 > [ 72.285132] [<ffffffff8129ef88>] __xfs_get_blocks+0x14c/0xf6c > [ 72.286456] [<ffffffff8103d74b>] ? get_user_pages_fast+0xee/0x14c > [ 72.287847] [<ffffffff8129fdcf>] xfs_get_blocks_direct+0x14/0x16 > [ 72.289246] [<ffffffff811879c6>] get_blocks+0x86/0x182 > [ 72.297467] [<ffffffff8103d74b>] ? get_user_pages_fast+0xee/0x14c > [ 72.298845] [<ffffffff81187fe6>] __blockdev_direct_IO+0x372/0x666 > [ 72.300266] [<ffffffff8129fdbb>] ? xfs_get_blocks+0x13/0x13 > [ 72.301555] [<ffffffff8116f9ac>] ? __mnt_drop_write_file+0x12/0x14 > [ 72.302936] [<ffffffff8116a017>] ? file_update_time+0xb5/0xcb > [ 72.304279] [<ffffffff8129d890>] xfs_vm_direct_IO+0x89/0x90 > [ 72.305555] [<ffffffff8129d943>] ? xfs_setfilesize_trans_alloc+0xac/0xac > [ 72.307064] [<ffffffff8169ae5e>] xfs_file_dio_aio_write+0x2e0/0x449 > [ 72.308509] [<ffffffff813e1028>] ? __clear_user+0x36/0x5b > [ 72.309755] [<ffffffff812af9c6>] xfs_file_write_iter+0x75/0x105 > [ 72.311107] [<ffffffff81151524>] __vfs_write+0x97/0xc0 > [ 72.312347] [<ffffffff81151b1f>] vfs_write+0xb5/0x16f > [ 72.313526] [<ffffffff811522c3>] SyS_write+0x4a/0x94 > [ 72.314683] [<ffffffff816a4b17>] system_call_fastpath+0x12/0x6f There should not be unwritten extent conversion happening through the allocation path here. Indeed, xfs_iomap_write_direct() only uses the XFS_BMAPI_PREALLOC flag, which means xfs_bmapi_convert_unwritten() returns without doing anything to the newly allocated extent... > [ 72.328096] Call Trace: > [ 72.328786] [<ffffffff8169c2f3>] dump_stack+0x4f/0x7b > [ 72.330002] [<ffffffff8127140f>] xfs_bmapi_convert_unwritten+0x73/0x176 > [ 72.331565] [<ffffffff8126cd9d>] ? xfs_bmap_search_extents+0x60/0xd6 > [ 72.333059] [<ffffffff812c96f1>] ? kmem_zone_alloc+0x6e/0xba > [ 72.334412] [<ffffffff8127a233>] xfs_bmapi_write+0x2a0/0x7ee > [ 72.335777] [<ffffffff812b9455>] xfs_iomap_write_unwritten+0x205/0x413 > [ 72.337300] [<ffffffff8129d7e2>] xfs_end_io+0x50/0x75 > [ 72.338495] [<ffffffff8129dab9>] xfs_end_io_direct_write+0x176/0x26c > [ 72.339952] [<ffffffff810d20a4>] ? delayacct_end+0x55/0x5e > [ 72.341238] [<ffffffff8118759d>] dio_complete+0x7c/0x134 > [ 72.342488] [<ffffffff811882a7>] __blockdev_direct_IO+0x633/0x666 > [ 72.343908] [<ffffffff8129fdbb>] ? xfs_get_blocks+0x13/0x13 > [ 72.345263] [<ffffffff8129d890>] xfs_vm_direct_IO+0x89/0x90 > [ 72.346598] [<ffffffff8129d943>] ? xfs_setfilesize_trans_alloc+0xac/0xac > [ 72.348185] [<ffffffff8169ae5e>] xfs_file_dio_aio_write+0x2e0/0x449 > [ 72.349652] [<ffffffff813e1028>] ? __clear_user+0x36/0x5b > [ 72.350950] [<ffffffff812af9c6>] xfs_file_write_iter+0x75/0x105 > [ 72.352369] [<ffffffff81151524>] __vfs_write+0x97/0xc0 > [ 72.353607] [<ffffffff81151b1f>] vfs_write+0xb5/0x16f > [ 72.354832] [<ffffffff811522c3>] SyS_write+0x4a/0x94 > [ 72.356060] [<ffffffff816a4b17>] system_call_fastpath+0x12/0x6f This is the only unwritten conversion path that should occur through direct IO - the IO completion path.... Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
