The patch titled Subject: ocfs2: refcount: take inode_lock until write io issued has been removed from the -mm tree. Its filename was refcount-take-inode_lock-until-write-io-issued.patch This patch was dropped because it was nacked ------------------------------------------------------ From: Wengang Wang <wen.gang.wang@xxxxxxxxxx> Subject: ocfs2: refcount: take inode_lock until write io issued This patch tries to fix this crash: #5 [ffff88003c1cd690] do_invalid_op at ffffffff810166d5 #6 [ffff88003c1cd730] invalid_op at ffffffff8159b2de [exception RIP: ocfs2_direct_IO_get_blocks+359] RIP: ffffffffa05dfa27 RSP: ffff88003c1cd7e8 RFLAGS: 00010202 RAX: 0000000000000000 RBX: ffff88003c1cdaa8 RCX: 0000000000000000 RDX: 000000000000000c RSI: ffff880027a95000 RDI: ffff88003c79b540 RBP: ffff88003c1cd858 R8: 0000000000000000 R9: ffffffff815f6ba0 R10: 00000000000001c9 R11: 00000000000001c9 R12: ffff88002d271500 R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000001000 ORIG_RAX: ffffffffffffffff CS: 0010 SS: 0018 #7 [ffff88003c1cd860] do_direct_IO at ffffffff811cd31b #8 [ffff88003c1cd950] direct_IO_iovec at ffffffff811cde9c #9 [ffff88003c1cd9b0] do_blockdev_direct_IO at ffffffff811ce764 #10 [ffff88003c1cdb80] __blockdev_direct_IO at ffffffff811ce7cc #11 [ffff88003c1cdbb0] ocfs2_direct_IO at ffffffffa05df756 [ocfs2] #12 [ffff88003c1cdbe0] generic_file_direct_write_iter at ffffffff8112f935 #13 [ffff88003c1cdc40] ocfs2_file_write_iter at ffffffffa0600ccc [ocfs2] #14 [ffff88003c1cdd50] do_aio_write at ffffffff8119126c #15 [ffff88003c1cddc0] aio_rw_vect_retry at ffffffff811d9bb4 #16 [ffff88003c1cddf0] aio_run_iocb at ffffffff811db880 #17 [ffff88003c1cde30] io_submit_one at ffffffff811dc238 #18 [ffff88003c1cde80] do_io_submit at ffffffff811dc437 #19 [ffff88003c1cdf70] sys_io_submit at ffffffff811dc530 #20 [ffff88003c1cdf80] system_call_fastpath at ffffffff8159a159 It crashes at BUG_ON(create && (ext_flags & OCFS2_EXT_REFCOUNTED)); in ocfs2_direct_IO_get_blocks. ocfs2_direct_IO_get_blocks is expecting the OCFS2_EXT_REFCOUNTED be removed in ocfs2_prepare_inode_for_write() if it was there. But no cluster lock is taken during the time before (or inside) ocfs2_prepare_inode_for_write() and after ocfs2_direct_IO_get_blocks(). It can happen in this case: Node A(which crashes) Node B ------------------------ --------------------------- ocfs2_file_aio_write ocfs2_prepare_inode_for_write ocfs2_inode_lock ... ocfs2_inode_unlock #no refcount found .... ocfs2_reflink ocfs2_inode_lock ... ocfs2_inode_unlock #now, refcount flag set on extent ... flush change to disk ocfs2_direct_IO_get_blocks ocfs2_get_clusters #extent map miss #buffer_head miss read extents from disk found refcount flag on extent crash.. Fix: We have to take the inode_lock long enough until ocfs2_direct_IO_get_blocks finished. Signed-off-by: Wengang Wang <wen.gang.wang@xxxxxxxxxx> Cc: Mark Fasheh <mfasheh@xxxxxxxx> Cc: Joel Becker <jlbec@xxxxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- fs/ocfs2/file.c | 20 ++++++++++++++++++-- fs/ocfs2/ocfs2_fs.h | 2 +- 2 files changed, 19 insertions(+), 3 deletions(-) diff -puN fs/ocfs2/file.c~refcount-take-inode_lock-until-write-io-issued fs/ocfs2/file.c --- a/fs/ocfs2/file.c~refcount-take-inode_lock-until-write-io-issued +++ a/fs/ocfs2/file.c @@ -2104,13 +2104,16 @@ static int ocfs2_prepare_inode_for_write size_t count, int appending, int *direct_io, - int *has_refcount) + int *has_refcount, + int *meta_level_out) { int ret = 0, meta_level = 0; struct dentry *dentry = file->f_path.dentry; struct inode *inode = dentry->d_inode; loff_t saved_pos = 0, end; + if (meta_level_out) + *meta_level_out = -1; /* * We start with a read level meta lock and only jump to an ex * if we need to make modifications here. @@ -2226,6 +2229,15 @@ out_unlock: saved_pos, appending, count, direct_io, has_refcount); + /* + * If direct IO would be done later, we have to keep inode_lock locked. + * Buffer'd IO is fine since the COW work will be done again in + * ocfs2_write_begin. + */ + if (direct_io && *direct_io && meta_level_out && !ret) { + *meta_level_out = meta_level; + meta_level = -1; + } if (meta_level >= 0) ocfs2_inode_unlock(inode, meta_level); @@ -2248,6 +2260,7 @@ static ssize_t ocfs2_file_write_iter(str int full_coherency = !(osb->s_mount_opt & OCFS2_MOUNT_COHERENCY_BUFFERED); int unaligned_dio = 0; + int meta_level = -1; trace_ocfs2_file_aio_write(inode, file, file->f_path.dentry, (unsigned long long)OCFS2_I(inode)->ip_blkno, @@ -2307,7 +2320,8 @@ relock: can_do_direct = direct_io; ret = ocfs2_prepare_inode_for_write(file, ppos, iocb->ki_nbytes, appending, - &can_do_direct, &has_refcount); + &can_do_direct, &has_refcount, + &meta_level); if (ret < 0) { mlog_errno(ret); goto out; @@ -2423,6 +2437,8 @@ out_sems: if (have_alloc_sem) ocfs2_iocb_clear_sem_locked(iocb); + if (meta_level >= 0) + ocfs2_inode_unlock(inode, meta_level); mutex_unlock(&inode->i_mutex); if (written) diff -puN fs/ocfs2/ocfs2_fs.h~refcount-take-inode_lock-until-write-io-issued fs/ocfs2/ocfs2_fs.h --- a/fs/ocfs2/ocfs2_fs.h~refcount-take-inode_lock-until-write-io-issued +++ a/fs/ocfs2/ocfs2_fs.h @@ -724,7 +724,7 @@ struct ocfs2_dinode { __le64 i_xattr_loc; /*80*/ struct ocfs2_block_check i_check; /* Error checking */ /*88*/ __le64 i_dx_root; /* Pointer to dir index root block */ -/*90*/ __le64 i_refcount_loc; +/*90*/ __le64 i_refcount_loc; /* Root block of the refcount tree */ __le64 i_suballoc_loc; /* Suballocator block group this inode belongs to. Only valid if allocated from a _ Patches currently in -mm which might be from wen.gang.wang@xxxxxxxxxx are -- To unsubscribe from this list: send the line "unsubscribe mm-commits" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html