Skip to contentSkip to conversation

Return more useful error number when acls are too large

PATCH In review Andreas Gruenbacher 3 participants 5 messages Apr 26, 2004 – Apr 27, 2004

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Return more useful error number when acls are too large

Hello,

could you please add this to mainline? Getting EINVAL when an acl becomes too large is quite confusing.

fs/ext2/acl.c +1 -1
Index: linux-2.6.6-rc2/fs/ext2/acl.c
===================================================================
@@ -256,7 +256,7 @@
 	}
  	if (acl) {
 		if (acl->a_count > EXT2_ACL_MAX_ENTRIES)
-			return -EINVAL;
+			return -ENOSPC;
 		value = ext2_acl_to_disk(acl, &size);
 		if (IS_ERR(value))
 			return (int)PTR_ERR(value);
fs/ext3/acl.c +1 -1
Index: linux-2.6.6-rc2/fs/ext3/acl.c
===================================================================
@@ -260,7 +260,7 @@
 	}
  	if (acl) {
 		if (acl->a_count > EXT3_ACL_MAX_ENTRIES)
-			return -EINVAL;
+			return -ENOSPC;
 		value = ext3_acl_to_disk(acl, &size);
 		if (IS_ERR(value))
 			return (int)PTR_ERR(value);

Thanks,

Signature · 2 lines
Andreas Gruenbacher <[email protected]>
SUSE Labs, SUSE LINUX AG
#1 lore raw

hi Andreas,

fs/ext2/acl.c line 259

Quoted 9 lines · Andreas Gruenbacher

Hello,

could you please add this to mainline? Getting EINVAL when an acl becomes too large is quite confusing.

  	if (acl) {
fs/ext2/acl.c +1 -1
 		if (acl->a_count > EXT2_ACL_MAX_ENTRIES)
-			return -EINVAL;
+			return -ENOSPC;

That seems an odd error code to change it to, since its not related to the device running out of free space right? Maybe use -E2BIG instead?

XFS has a similar check (different limit as you know), so is also affected by this; could you update XFS at the same time with whatever value gets chosen, if its not EINVAL? Or just let me know what gets chosen & I'll fix it up later.

thanks.

Signature · 1 line
Nathan
#2 lore raw
Quoted 8 lines · Nathan Scott

hi Andreas,

Quoted 5 lines · Andreas Gruenbacher

Hello,

could you please add this to mainline? Getting EINVAL when an acl becomes too large is quite confusing.

Quoted 8 lines

Review

Quoted 4 lines
  	if (acl) {
(quoted diff) +1 -1
 		if (acl->a_count > EXT2_ACL_MAX_ENTRIES)
-			return -EINVAL;
+			return -ENOSPC;

That seems an odd error code to change it to, since its not related to the device running out of free space right? Maybe use -E2BIG instead?

I don't see a problem with giving ENOSPC this particular meaning here. The error number used must be among those defined for NFSv3, so E2BIG won't do.

Quoted 4 lines

XFS has a similar check (different limit as you know), so is also affected by this; could you update XFS at the same time with whatever value gets chosen, if its not EINVAL? Or just let me know what gets chosen & I'll fix it up later.

I think this patch is correct for xfs. Nathan, would you mind to double-check? Thanks.

fs/xfs/xfs_acl.c +4 -8
Index: linux-2.6.6-rc2/fs/xfs/xfs_acl.c
===================================================================
@@ -148,10 +148,7 @@ posix_acl_xattr_to_xfs(
 			return EINVAL;
 		}
 	}
-	if (xfs_acl_invalid(dest))
-		return EINVAL;
-
-	return 0;
+	return xfs_acl_invalid(dest);
 }
 
 /*
@@ -249,10 +246,9 @@ xfs_acl_vget(
 	if (!size) {
 		error = -posix_acl_xattr_size(XFS_ACL_MAX_ENTRIES);
 	} else {
-		if (xfs_acl_invalid(xfs_acl)) {
-			error = EINVAL;
+		error = xfs_acl_invalid(xfs_acl);
+		if (error)
 			goto out;
-		}
 		if (kind == _ACL_TYPE_ACCESS) {
 			vattr_t	va;
 
@@ -590,7 +586,7 @@ xfs_acl_invalid(
 		goto acl_invalid;
 
 	if (aclp->acl_cnt > XFS_ACL_MAX_ENTRIES)
-		goto acl_invalid;
+		return ENOSPC;
 
 	for (i = 0; i < aclp->acl_cnt; i++) {
 		entry = &aclp->acl_entry[i];

Thanks,

Signature · 2 lines
Andreas Gruenbacher <[email protected]>
SUSE Labs, SUSE LINUX AG
fs/xfs/xfs_acl.c +4 -8
Index: linux-2.6.6-rc2/fs/xfs/xfs_acl.c
===================================================================
@@ -148,10 +148,7 @@ posix_acl_xattr_to_xfs(
 			return EINVAL;
 		}
 	}
-	if (xfs_acl_invalid(dest))
-		return EINVAL;
-
-	return 0;
+	return xfs_acl_invalid(dest);
 }
 
 /*
@@ -249,10 +246,9 @@ xfs_acl_vget(
 	if (!size) {
 		error = -posix_acl_xattr_size(XFS_ACL_MAX_ENTRIES);
 	} else {
-		if (xfs_acl_invalid(xfs_acl)) {
-			error = EINVAL;
+		error = xfs_acl_invalid(xfs_acl);
+		if (error)
 			goto out;
-		}
 		if (kind == _ACL_TYPE_ACCESS) {
 			vattr_t	va;
 
@@ -590,7 +586,7 @@ xfs_acl_invalid(
 		goto acl_invalid;
 
 	if (aclp->acl_cnt > XFS_ACL_MAX_ENTRIES)
-		goto acl_invalid;
+		return ENOSPC;
 
 	for (i = 0; i < aclp->acl_cnt; i++) {
 		entry = &aclp->acl_entry[i];
#3 lore raw
In reply to Andreas Gruenbacher
Quoted 2 lines · Andreas Gruenbacher

could you please add this to mainline? Getting EINVAL when an acl becomes too large is quite confusing.

On my system, at least, "man acl_set_file" does explicitly say that EINVAL is returned in this case. Whether that should be considered a bug in the documentation or the code I don't know....

--Bruce Fields

#4 lore raw
Quoted 7 lines · J. Bruce Fields
Quoted 2 lines · Andreas Gruenbacher

could you please add this to mainline? Getting EINVAL when an acl becomes too large is quite confusing.

On my system, at least, "man acl_set_file" does explicitly say that EINVAL is returned in this case. Whether that should be considered a bug in the documentation or the code I don't know....

Indeed ... looking at POSIX 1003.1e draft 17 I now see that this function is indeed supposed to return EINVAL in this case. I was assuming that this case was undefined, but that was wrong. So let's leave the code as is -- sorry.

Cheers,

Signature · 2 lines
Andreas Gruenbacher <[email protected]>
SUSE Labs, SUSE LINUX AG