Skip to contentSkip to conversation

xfs: remove incorrect ASSERT in xfs_rename

PATCH In review Kuniyuki Iwashima 3 participants 7 messages Jul 7, 2022 – Jul 8, 2022 Stable branch

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xfs: remove incorrect ASSERT in xfs_rename

From: Eric Sandeen <[email protected]>

commit e445976537ad139162980bee015b7364e5b64fff upstream.

Ayushman Dutta reported our 5.10 kernel hit the warning. It was because the original commit misses a Fixes tag and was not backported to the stable tree. The fix is merged in 5.16, so please backport it to 5.15 first.

This ASSERT in xfs_rename is a) incorrect, because (RENAME_WHITEOUT|RENAME_NOREPLACE) is a valid combination, and
b) unnecessary, because actual invalid flag combinations are already handled at the vfs level in do_renameat2() before we get called. So, remove it.

---
I will send another patch for 4.9 - 5.4 because of a conflict with idmapped mount changes.

---
 fs/xfs/xfs_inode.c | 1 -
 1 file changed, 1 deletion(-)
fs/xfs/xfs_inode.c +0 -1
index 2477e301fa82..c19f3ca605af 100644
@@ -3128,7 +3128,6 @@ xfs_rename(
 	 * appropriately.
 	 */
 	if (flags & RENAME_WHITEOUT) {
-		ASSERT(!(flags & (RENAME_NOREPLACE | RENAME_EXCHANGE)));
 		error = xfs_rename_alloc_whiteout(mnt_userns, target_dp, &wip);
 		if (error)
 			return error;
Signature · 1 line
2.30.2
Acked-by on v1 1/1
#1 lore raw
Quoted 14 lines · Kuniyuki Iwashima

From: Eric Sandeen <[email protected]>

commit e445976537ad139162980bee015b7364e5b64fff upstream.

Ayushman Dutta reported our 5.10 kernel hit the warning. It was because the original commit misses a Fixes tag and was not backported to the stable tree. The fix is merged in 5.16, so please backport it to 5.15 first.

This ASSERT in xfs_rename is a) incorrect, because (RENAME_WHITEOUT|RENAME_NOREPLACE) is a valid combination, and
b) unnecessary, because actual invalid flag combinations are already handled at the vfs level in do_renameat2() before we get called. So, remove it.

Quoted 7 lines

Looks good to me, but you really ought to send 5.10 patches to the 5.10 XFS maintainer (Amir, now cc'd). (Yes, this is a recent change.) ;)

--D

Quoted 22 lines

---
I will send another patch for 4.9 - 5.4 because of a conflict with idmapped mount changes.

---
 fs/xfs/xfs_inode.c | 1 -
 1 file changed, 1 deletion(-)
fs/xfs/xfs_inode.c +0 -1
index 2477e301fa82..c19f3ca605af 100644
@@ -3128,7 +3128,6 @@ xfs_rename(
 	 * appropriately.
 	 */
 	if (flags & RENAME_WHITEOUT) {
-		ASSERT(!(flags & (RENAME_NOREPLACE | RENAME_EXCHANGE)));
 		error = xfs_rename_alloc_whiteout(mnt_userns, target_dp, &wip);
 		if (error)
 			return error;
Signature · 1 line
2.30.2
#2 lore raw
Quoted 23 lines · Darrick J. Wong
Quoted 14 lines · Kuniyuki Iwashima

From: Eric Sandeen <[email protected]>

commit e445976537ad139162980bee015b7364e5b64fff upstream.

Ayushman Dutta reported our 5.10 kernel hit the warning. It was because the original commit misses a Fixes tag and was not backported to the stable tree. The fix is merged in 5.16, so please backport it to 5.15 first.

This ASSERT in xfs_rename is a) incorrect, because (RENAME_WHITEOUT|RENAME_NOREPLACE) is a valid combination, and
b) unnecessary, because actual invalid flag combinations are already handled at the vfs level in do_renameat2() before we get called. So, remove it.

Quoted 7 lines
Quoted 2 lines

Looks good to me, but you really ought to send 5.10 patches to the 5.10 XFS maintainer (Amir, now cc'd). (Yes, this is a recent change.) ;)

...and of course the first thing that happens is that I mix up the 5.10 and 5.15 patches.

Amir is the 5.10 maintainer, Leah is the 5.15 maintainer. Sorry about the mixup. /me pours himself a third(!) cup of coffee.

--D

Quoted 26 lines

--D

Quoted 22 lines

---
I will send another patch for 4.9 - 5.4 because of a conflict with idmapped mount changes.

---
 fs/xfs/xfs_inode.c | 1 -
 1 file changed, 1 deletion(-)
fs/xfs/xfs_inode.c +0 -1
index 2477e301fa82..c19f3ca605af 100644
@@ -3128,7 +3128,6 @@ xfs_rename(
 	 * appropriately.
 	 */
 	if (flags & RENAME_WHITEOUT) {
-		ASSERT(!(flags & (RENAME_NOREPLACE | RENAME_EXCHANGE)));
 		error = xfs_rename_alloc_whiteout(mnt_userns, target_dp, &wip);
 		if (error)
 			return error;
Signature · 1 line
2.30.2
#3 lore raw
From:   "Darrick J. Wong" <[email protected]>
Date:   Fri, 8 Jul 2022 08:55:57 -0700
Quoted 24 lines · Darrick J. Wong
Quoted 23 lines · Kuniyuki Iwashima
Quoted 21 lines

From: Eric Sandeen <[email protected]>

commit e445976537ad139162980bee015b7364e5b64fff upstream.

Ayushman Dutta reported our 5.10 kernel hit the warning. It was because the original commit misses a Fixes tag and was not backported to the stable tree. The fix is merged in 5.16, so please backport it to 5.15 first.

This ASSERT in xfs_rename is a) incorrect, because (RENAME_WHITEOUT|RENAME_NOREPLACE) is a valid combination, and
b) unnecessary, because actual invalid flag combinations are already handled at the vfs level in do_renameat2() before we get called. So, remove it.

Quoted 8 lines
Quoted 2 lines

Looks good to me, but you really ought to send 5.10 patches to the 5.10 XFS maintainer (Amir, now cc'd). (Yes, this is a recent change.) ;)

...and of course the first thing that happens is that I mix up the 5.10 and 5.15 patches.

Amir is the 5.10 maintainer, Leah is the 5.15 maintainer. Sorry about the mixup. /me pours himself a third(!) cup of coffee.

Thank you for taking a look!

And sorry that I'm not familiar with xfs workflow and didn't know each version has dedicated maintainers.

Is there a doc like Documentation/process/maintainer-netdev.rst as both of Amir and Leah seem not listed in the xfs entry of MAINTAINERS...?

#4 lore raw
Quoted 32 lines · Kuniyuki Iwashima
From:   "Darrick J. Wong" <[email protected]>
Date:   Fri, 8 Jul 2022 08:55:57 -0700
Quoted 30 lines · Darrick J. Wong
Quoted 23 lines · Kuniyuki Iwashima
Quoted 14 lines

From: Eric Sandeen <[email protected]>

commit e445976537ad139162980bee015b7364e5b64fff upstream.

Ayushman Dutta reported our 5.10 kernel hit the warning. It was because the original commit misses a Fixes tag and was not backported to the stable tree. The fix is merged in 5.16, so please backport it to 5.15 first.

This ASSERT in xfs_rename is a) incorrect, because (RENAME_WHITEOUT|RENAME_NOREPLACE) is a valid combination, and
b) unnecessary, because actual invalid flag combinations are already handled at the vfs level in do_renameat2() before we get called. So, remove it.

Quoted 7 lines
Quoted 2 lines

Looks good to me, but you really ought to send 5.10 patches to the 5.10 XFS maintainer (Amir, now cc'd). (Yes, this is a recent change.) ;)

...and of course the first thing that happens is that I mix up the 5.10 and 5.15 patches.

Quoted 7 lines
Quoted 2 lines

Amir is the 5.10 maintainer, Leah is the 5.15 maintainer. Sorry about the mixup. /me pours himself a third(!) cup of coffee.

Thank you for taking a look!

And sorry that I'm not familiar with xfs workflow and didn't know each version has dedicated maintainers.

It's a recent change, because I wasn't keeping up with tending to six LTS trees /and/ upstream /and/ feature development.

Quoted 2 lines

Is there a doc like Documentation/process/maintainer-netdev.rst as both of Amir and Leah seem not listed in the xfs entry of MAINTAINERS...?

They're listed in MAINTAINERS in the 5.10 and 5.15 trees, respectively. That's also a very recent change (within the last week, I think).

--D

#5 lore raw
From:   "Darrick J. Wong" <[email protected]>
Date:   Fri, 8 Jul 2022 09:51:56 -0700
Quoted 36 lines · Kuniyuki Iwashima
Quoted 35 lines
From:   "Darrick J. Wong" <[email protected]>
Date:   Fri, 8 Jul 2022 08:55:57 -0700
Quoted 32 lines · Darrick J. Wong
Quoted 23 lines · Kuniyuki Iwashima
Quoted 14 lines

From: Eric Sandeen <[email protected]>

commit e445976537ad139162980bee015b7364e5b64fff upstream.

Ayushman Dutta reported our 5.10 kernel hit the warning. It was because the original commit misses a Fixes tag and was not backported to the stable tree. The fix is merged in 5.16, so please backport it to 5.15 first.

This ASSERT in xfs_rename is a) incorrect, because (RENAME_WHITEOUT|RENAME_NOREPLACE) is a valid combination, and
b) unnecessary, because actual invalid flag combinations are already handled at the vfs level in do_renameat2() before we get called. So, remove it.

Quoted 7 lines
Quoted 2 lines

Looks good to me, but you really ought to send 5.10 patches to the 5.10 XFS maintainer (Amir, now cc'd). (Yes, this is a recent change.) ;)

...and of course the first thing that happens is that I mix up the 5.10 and 5.15 patches.

Amir is the 5.10 maintainer, Leah is the 5.15 maintainer. Sorry about the mixup. /me pours himself a third(!) cup of coffee.

Quoted 7 lines
Quoted 4 lines

Thank you for taking a look!

And sorry that I'm not familiar with xfs workflow and didn't know each version has dedicated maintainers.

It's a recent change, because I wasn't keeping up with tending to six LTS trees /and/ upstream /and/ feature development.

It must have been really hard, that makes sense.

Quoted 5 lines
Quoted 2 lines

Is there a doc like Documentation/process/maintainer-netdev.rst as both of Amir and Leah seem not listed in the xfs entry of MAINTAINERS...?

They're listed in MAINTAINERS in the 5.10 and 5.15 trees, respectively. That's also a very recent change (within the last week, I think).

Ah, I got it. I'm sorry it was because I used the latest get_maintainer.pl only. I'll check each branch's MAINTAINERS next time.

Thank you!

#6 lore raw
In reply to Darrick J. Wong
Quoted 23 lines · Darrick J. Wong
Quoted 14 lines · Kuniyuki Iwashima

From: Eric Sandeen <[email protected]>

commit e445976537ad139162980bee015b7364e5b64fff upstream.

Ayushman Dutta reported our 5.10 kernel hit the warning. It was because the original commit misses a Fixes tag and was not backported to the stable tree. The fix is merged in 5.16, so please backport it to 5.15 first.

This ASSERT in xfs_rename is a) incorrect, because (RENAME_WHITEOUT|RENAME_NOREPLACE) is a valid combination, and
b) unnecessary, because actual invalid flag combinations are already handled at the vfs level in do_renameat2() before we get called. So, remove it.

Quoted 7 lines
Quoted 6 lines

Looks good to me, but you really ought to send 5.10 patches to the 5.10 XFS maintainer (Amir, now cc'd). (Yes, this is a recent change.) ;)

--D

This patch is actually part of the next set of 10 patches being testing for the 5.15 branch :) It would have been going out in the next week or two, but since this is such a minor change, we can just go ahead with it.

- Leah

Quoted 23 lines
Quoted 23 lines

---
I will send another patch for 4.9 - 5.4 because of a conflict with idmapped mount changes.

---
 fs/xfs/xfs_inode.c | 1 -
 1 file changed, 1 deletion(-)
fs/xfs/xfs_inode.c +0 -1
index 2477e301fa82..c19f3ca605af 100644
@@ -3128,7 +3128,6 @@ xfs_rename(
 	 * appropriately.
 	 */
 	if (flags & RENAME_WHITEOUT) {
-		ASSERT(!(flags & (RENAME_NOREPLACE | RENAME_EXCHANGE)));
 		error = xfs_rename_alloc_whiteout(mnt_userns, target_dp, &wip);
 		if (error)
 			return error;
Signature · 1 line
2.30.2