Re: [PATCH] ubifs: Fix inode leak in xattr code

2017-05-16 Thread Richard Weinberger
Artem, Am 15.05.2017 um 18:05 schrieb Artem Bityutskiy: > On Mon, 2017-05-15 at 17:22 +0200, Richard Weinberger wrote: >> Alternatively we could add a iget_locked/drop_nlink/iput sequence to >> ubifs_tnc_remove_ino(). But that will make unlink() much slower for >> files >> that contain xattrs. >

Re: [PATCH] ubifs: Fix inode leak in xattr code

2017-05-15 Thread Richard Weinberger
Artem, Am 15.05.2017 um 18:05 schrieb Artem Bityutskiy: > On Mon, 2017-05-15 at 17:22 +0200, Richard Weinberger wrote: >> Alternatively we could add a iget_locked/drop_nlink/iput sequence to >> ubifs_tnc_remove_ino(). But that will make unlink() much slower for >> files >> that contain xattrs. >

Re: [PATCH] ubifs: Fix inode leak in xattr code

2017-05-15 Thread Richard Weinberger
Artem, Am 15.05.2017 um 18:05 schrieb Artem Bityutskiy: > On Mon, 2017-05-15 at 17:22 +0200, Richard Weinberger wrote: >> Alternatively we could add a iget_locked/drop_nlink/iput sequence to >> ubifs_tnc_remove_ino(). But that will make unlink() much slower for >> files >> that contain xattrs. >

Re: [PATCH] ubifs: Fix inode leak in xattr code

2017-05-15 Thread Artem Bityutskiy
On Mon, 2017-05-15 at 17:22 +0200, Richard Weinberger wrote: > Alternatively we could add a iget_locked/drop_nlink/iput sequence to > ubifs_tnc_remove_ino(). But that will make unlink() much slower for > files > that contain xattrs. At that level we'd need to do it for every xattr, even those that

Re: [PATCH] ubifs: Fix inode leak in xattr code

2017-05-15 Thread Richard Weinberger
Artem, Am 15.05.2017 um 16:53 schrieb Artem Bityutskiy: > On Mon, 2017-05-15 at 16:20 +0200, Richard Weinberger wrote: >> To solve this problem, set i_nlink for all xattr inodes to 0, such >> that >> the iput() in the UBIFS xattr code makes the temporary inode vanish >> immediately. > > What if t

Re: [PATCH] ubifs: Fix inode leak in xattr code

2017-05-15 Thread Artem Bityutskiy
On Mon, 2017-05-15 at 16:20 +0200, Richard Weinberger wrote: > To solve this problem, set i_nlink for all xattr inodes to 0, such > that > the iput() in the UBIFS xattr code makes the temporary inode vanish > immediately. What if there is iget right after iput? With this patch, will we need to go

[PATCH] ubifs: Fix inode leak in xattr code

2017-05-15 Thread Richard Weinberger
UBIFS handles extended attributes just like files, as consequence of that, they also have inodes. Therefore UBIFS does all the inode machinery also for xattrs. Since new inodes have i_nlink of 1, a file or xattr inode will be evicted if i_nlink goes down to 0 after an unlink. UBIFS assumes this mod