Skip to contentSkip to conversation

inotify: hide internal kernel bits from fdinfo

PATCH In review Dave Hansen 4 participants 4 messages Sep 21, 2015 Fsnotify: filesystem notification infrastructure

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 inotify: hide internal kernel bits from fdinfo

From: Dave Hansen <[email protected]>

There was a report that my patch:

inotify: actually check for invalid bits in sys_inotify_add_watch()

broke CRIU.

The reason is that CRIU looks up raw flags in /proc/$pid/fdinfo/* to figure out how to rebuild inotify watches and then passes those flags directly back in to the inotify API. One of those flags (FS_EVENT_ON_CHILD) is set in mark->mask, but is not part of the inotify API. It is used inside the kernel to _implement_ inotify but it is not and has never been part of the API.

My patch above ensured that we only allow bits which are part of the API (IN_ALL_EVENTS). This broke CRIU.

FS_EVENT_ON_CHILD is really internal to the kernel. It is set _anyway_ on all inotify marks. So, CRIU was really just trying to set a bit that was already set.

This patch hides that bit from fdinfo. CRIU will not see the bit, not try to set it, and should work as before. We should not have been exposing this bit in the first place, so this is a good patch independent of the CRIU problem.

---

 b/fs/notify/fdinfo.c |    9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
fs/notify/fdinfo.c renamed from fs/notify/fdinfo.c~fdinfo-mask +8 -1
diff -puN fs/notify/fdinfo.c~fdinfo-mask fs/notify/fdinfo.c
@@ -82,9 +82,16 @@ static void inotify_fdinfo(struct seq_fi
 	inode_mark = container_of(mark, struct inotify_inode_mark, fsn_mark);
 	inode = igrab(mark->inode);
 	if (inode) {
+		/*
+		 * IN_ALL_EVENTS represents all of the mask bits
+		 * that we expose to userspace.  There is at
+		 * least one bit (FS_EVENT_ON_CHILD) which is
+		 * used only internally to the kernel.
+		 */
+		u32 mask = mark->mask & IN_ALL_EVENTS;
 		seq_printf(m, "inotify wd:%x ino:%lx sdev:%x mask:%x ignored_mask:%x ",
 			   inode_mark->wd, inode->i_ino, inode->i_sb->s_dev,
-			   mark->mask, mark->ignored_mask);
+			   mask, mark->ignored_mask);
 		show_mark_fhandle(m, inode);
 		seq_putc(m, '\n');
 		iput(inode);

_

Acked-by on v1 1/1
#1 lore raw
Quoted 68 lines · Dave Hansen

From: Dave Hansen <[email protected]>

There was a report that my patch:

inotify: actually check for invalid bits in
sys_inotify_add_watch()

broke CRIU.

The reason is that CRIU looks up raw flags in /proc/$pid/fdinfo/* to figure out how to rebuild inotify watches and then passes those flags directly back in to the inotify API. One of those flags (FS_EVENT_ON_CHILD) is set in mark->mask, but is not part of the inotify API. It is used inside the kernel to _implement_ inotify but it is not and has never been part of the API.

My patch above ensured that we only allow bits which are part of the API (IN_ALL_EVENTS). This broke CRIU.

FS_EVENT_ON_CHILD is really internal to the kernel. It is set _anyway_ on all inotify marks. So, CRIU was really just trying to set a bit that was already set.

This patch hides that bit from fdinfo. CRIU will not see the bit, not try to set it, and should work as before. We should not have been exposing this bit in the first place, so this is a good patch independent of the CRIU problem.

---

 b/fs/notify/fdinfo.c |    9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff -puN fs/notify/fdinfo.c~fdinfo-mask fs/notify/fdinfo.c
--- a/fs/notify/fdinfo.c~fdinfo-mask 2015-09-21
10:24:01.031864268 -0700
+++ b/fs/notify/fdinfo.c 2015-09-21 10:25:04.335723826 -0700

fs/notify/fdinfo.c renamed from fs/notify/fdinfo.c~fdinfo-mask +8 -2
@@ -82,9 +82,16 @@ static void inotify_fdinfo(struct seq_fi
 	inode_mark = container_of(mark, struct inotify_inode_mark,
 fsn_mark);
 	inode = igrab(mark->inode);
 	if (inode) {
+		/*
+		 * IN_ALL_EVENTS represents all of the mask bits
+		 * that we expose to userspace.  There is at
+		 * least one bit (FS_EVENT_ON_CHILD) which is
+		 * used only internally to the kernel.
+		 */
+		u32 mask = mark->mask & IN_ALL_EVENTS;
 		seq_printf(m, "inotify wd:%x ino:%lx sdev:%x mask:%x
 ignored_mask:%x ",
 			   inode_mark->wd, inode->i_ino, inode->i_sb
->s_dev,
-			   mark->mask, mark->ignored_mask);
+			   mask, mark->ignored_mask);
 		show_mark_fhandle(m, inode);
 		seq_putc(m, '\n');
 		iput(inode);

_

Acked-by on v1 1/1
#2 lore raw
In reply to Dave Hansen
Quoted 23 lines · Dave Hansen

From: Dave Hansen <[email protected]>

There was a report that my patch:

inotify: actually check for invalid bits in sys_inotify_add_watch()

broke CRIU.

The reason is that CRIU looks up raw flags in /proc/$pid/fdinfo/* to figure out how to rebuild inotify watches and then passes those flags directly back in to the inotify API. One of those flags (FS_EVENT_ON_CHILD) is set in mark->mask, but is not part of the inotify API. It is used inside the kernel to _implement_ inotify but it is not and has never been part of the API.

My patch above ensured that we only allow bits which are part of the API (IN_ALL_EVENTS). This broke CRIU.

FS_EVENT_ON_CHILD is really internal to the kernel. It is set _anyway_ on all inotify marks. So, CRIU was really just trying to set a bit that was already set.

Quoted 7 lines

This patch hides that bit from fdinfo. CRIU will not see the bit, not try to set it, and should work as before. We should not have been exposing this bit in the first place, so this is a good patch independent of the CRIU problem.

Thank you!

Acked-by Andrey Vagin on v1 1/1
#3 lore raw
In reply to Dave Hansen

2015-09-21 21:45 GMT+03:00 Dave Hansen <[email protected]>:

Quoted 23 lines

From: Dave Hansen <[email protected]>

There was a report that my patch:

inotify: actually check for invalid bits in sys_inotify_add_watch()

broke CRIU.

The reason is that CRIU looks up raw flags in /proc/$pid/fdinfo/* to figure out how to rebuild inotify watches and then passes those flags directly back in to the inotify API. One of those flags (FS_EVENT_ON_CHILD) is set in mark->mask, but is not part of the inotify API. It is used inside the kernel to _implement_ inotify but it is not and has never been part of the API.

My patch above ensured that we only allow bits which are part of the API (IN_ALL_EVENTS). This broke CRIU.

FS_EVENT_ON_CHILD is really internal to the kernel. It is set _anyway_ on all inotify marks. So, CRIU was really just trying to set a bit that was already set.

Quoted 7 lines

This patch hides that bit from fdinfo. CRIU will not see the bit, not try to set it, and should work as before. We should not have been exposing this bit in the first place, so this is a good patch independent of the CRIU problem.

Thanks,
Andrey

Quoted 8 lines

---