The write context should also be freed even when direct IO failed. Otherwise a memory leak is introduced and entries remain in oi->ip_unwritten_list causing the following BUG later in unlink path:
ERROR: bug expression: !list_empty(&oi->ip_unwritten_list) ERROR: Clear inode of 215043, inode has unwritten extents ... Call Trace: ? __set_current_blocked+0x42/0x68 ocfs2_evict_inode+0x91/0x6a0 [ocfs2] ? bit_waitqueue+0x40/0x33 evict+0xdb/0x1af iput+0x1a2/0x1f7 do_unlinkat+0x194/0x28f SyS_unlinkat+0x1b/0x2f do_syscall_64+0x79/0x1ae entry_SYSCALL_64_after_hwframe+0x151/0x0
This patch also logs, with frequency limit, direct IO failures.
Signed-off-by: Wengang Wang wen.gang.wang@oracle.com --- v1 -> v2: log direct IO failures --- fs/ocfs2/aops.c | 12 ++++++++++-- fs/ocfs2/cluster/masklog.h | 9 +++++++++ 2 files changed, 19 insertions(+), 2 deletions(-)
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 302cd7c..7578bd5 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -2412,8 +2412,16 @@ static int ocfs2_dio_end_io(struct kiocb *iocb, /* this io's submitter should not have unlocked this before we could */ BUG_ON(!ocfs2_iocb_is_rw_locked(iocb));
- if (bytes > 0 && private) - ret = ocfs2_dio_end_io_write(inode, private, offset, bytes); + if (bytes <= 0) + mlog_ratelimited(ML_ERROR, "Direct IO failed, bytes = %lld", + (long long)bytes); + if (private) { + if (bytes > 0) + ret = ocfs2_dio_end_io_write(inode, private, offset, + bytes); + else + ocfs2_dio_free_write_ctx(inode, private); + }
ocfs2_iocb_clear_rw_locked(iocb);
diff --git a/fs/ocfs2/cluster/masklog.h b/fs/ocfs2/cluster/masklog.h index 308ea0e..a396096 100644 --- a/fs/ocfs2/cluster/masklog.h +++ b/fs/ocfs2/cluster/masklog.h @@ -178,6 +178,15 @@ do { \ ##__VA_ARGS__); \ } while (0)
+#define mlog_ratelimited(mask, fmt, ...) \ +do { \ + static DEFINE_RATELIMIT_STATE(_rs, \ + DEFAULT_RATELIMIT_INTERVAL, \ + DEFAULT_RATELIMIT_BURST); \ + if (__ratelimit(&_rs)) \ + mlog(mask, fmt, ##__VA_ARGS__); \ +} while (0) + #define mlog_errno(st) ({ \ int _st = (st); \ if (_st != -ERESTARTSYS && _st != -EINTR && \
Hi Guys,
Thanks for your reviews. As suggested by Junxiao, logging direct IO failures is added in v2. Please review the v2.
thanks, wengang
On 2018/11/2 10:06, Wengang Wang wrote:
The write context should also be freed even when direct IO failed. Otherwise a memory leak is introduced and entries remain in oi->ip_unwritten_list causing the following BUG later in unlink path:
ERROR: bug expression: !list_empty(&oi->ip_unwritten_list) ERROR: Clear inode of 215043, inode has unwritten extents ... Call Trace: ? __set_current_blocked+0x42/0x68 ocfs2_evict_inode+0x91/0x6a0 [ocfs2] ? bit_waitqueue+0x40/0x33 evict+0xdb/0x1af iput+0x1a2/0x1f7 do_unlinkat+0x194/0x28f SyS_unlinkat+0x1b/0x2f do_syscall_64+0x79/0x1ae entry_SYSCALL_64_after_hwframe+0x151/0x0
This patch also logs, with frequency limit, direct IO failures.
Signed-off-by: Wengang Wang wen.gang.wang@oracle.com
v1 -> v2: log direct IO failures
fs/ocfs2/aops.c | 12 ++++++++++-- fs/ocfs2/cluster/masklog.h | 9 +++++++++ 2 files changed, 19 insertions(+), 2 deletions(-)
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 302cd7c..7578bd5 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -2412,8 +2412,16 @@ static int ocfs2_dio_end_io(struct kiocb *iocb, /* this io's submitter should not have unlocked this before we could */ BUG_ON(!ocfs2_iocb_is_rw_locked(iocb));
- if (bytes > 0 && private)
ret = ocfs2_dio_end_io_write(inode, private, offset, bytes);
- if (bytes <= 0)
mlog_ratelimited(ML_ERROR, "Direct IO failed, bytes = %lld",(long long)bytes);- if (private) {
if (bytes > 0)ret = ocfs2_dio_end_io_write(inode, private, offset,bytes);elseocfs2_dio_free_write_ctx(inode, private);- }
ocfs2_iocb_clear_rw_locked(iocb); diff --git a/fs/ocfs2/cluster/masklog.h b/fs/ocfs2/cluster/masklog.h index 308ea0e..a396096 100644 --- a/fs/ocfs2/cluster/masklog.h +++ b/fs/ocfs2/cluster/masklog.h @@ -178,6 +178,15 @@ do { \ ##__VA_ARGS__); \ } while (0) +#define mlog_ratelimited(mask, fmt, ...) \ +do { \
- static DEFINE_RATELIMIT_STATE(_rs, \
DEFAULT_RATELIMIT_INTERVAL, \DEFAULT_RATELIMIT_BURST); \- if (__ratelimit(&_rs)) \
mlog(mask, fmt, ##__VA_ARGS__); \+} while (0)
- #define mlog_errno(st) ({ \ int _st = (st); \ if (_st != -ERESTARTSYS && _st != -EINTR && \
On Fri, Nov 02, 2018 at 10:06:32AM -0700, Wengang Wang wrote:
The write context should also be freed even when direct IO failed. Otherwise a memory leak is introduced and entries remain in oi->ip_unwritten_list causing the following BUG later in unlink path:
ERROR: bug expression: !list_empty(&oi->ip_unwritten_list) ERROR: Clear inode of 215043, inode has unwritten extents ... Call Trace: ? __set_current_blocked+0x42/0x68 ocfs2_evict_inode+0x91/0x6a0 [ocfs2] ? bit_waitqueue+0x40/0x33 evict+0xdb/0x1af iput+0x1a2/0x1f7 do_unlinkat+0x194/0x28f SyS_unlinkat+0x1b/0x2f do_syscall_64+0x79/0x1ae entry_SYSCALL_64_after_hwframe+0x151/0x0
This patch also logs, with frequency limit, direct IO failures.
Signed-off-by: Wengang Wang wen.gang.wang@oracle.com
v1 -> v2: log direct IO failures
fs/ocfs2/aops.c | 12 ++++++++++-- fs/ocfs2/cluster/masklog.h | 9 +++++++++ 2 files changed, 19 insertions(+), 2 deletions(-)
<formletter>
This is not the correct way to submit patches for inclusion in the stable kernel tree. Please read: https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html for how to do this properly.
</formletter>
linux-stable-mirror@lists.linaro.org
-
Greg KH -
Wengang Wang