On Sat, Dec 17, 2011 at 09:41:37PM +0000, Al Viro wrote: > xfs and ext4_ioctl() need to be fixed; XFS fix follows, ext4 I'd rather left > to ext4 folks - I don't know how wide an area needs i_mutex there Oh, for fsck sake... People, this is *obviously* broken - if nothing else, removing suid after modifying the file contents is too late. Moreover, this mext_inode_double_lock() thing is asking for trouble; it's deadlock-free only because nothing else takes i_mutex on more than one non-directory inode and does that as the innermost lock. Start calling it for directories (or have somebody cut'n'paste it and use it for directories) and you've got a nice, shiny deadlock... BTW, is ordering really needed in double_down_write_data_sem()? IOW, can we get contention between several callers of that thing? >From my reading of that code, all call chains leading to this sucker are guaranteed to already hold i_mutex on both inodes. If that is true, we don't need any ordering in double_down_write_data_sem() at all... AFAICS, the minimal fix is to move file_remove_suid() call into ext4_move_extents(), just after we have acquired i_mutex in there. Moreover, I think it should be done to *both* files, since both have contents modified. And I see no point in making that conditional... -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>