lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening PHC | |
Open Source and information security mailing list archives
| ||
|
Date: Thu, 16 Mar 2023 01:07:40 -0400 From: "Theodore Ts'o" <tytso@....edu> To: Kemeng Shi <shikemeng@...weicloud.com> Cc: adilger.kernel@...ger.ca, ojaswin@...ux.ibm.com, ritesh.list@...il.com, Harshad Shirwadkar <harshadshirwadkar@...il.com>, linux-ext4@...r.kernel.org, linux-kernel@...r.kernel.org Subject: Re: [PATCH v3 20/20] ext4: simplify calculation of blkoff in ext4_mb_new_blocks_simple On Sat, Mar 04, 2023 at 01:21:20AM +0800, Kemeng Shi wrote: > We try to allocate a block from goal in ext4_mb_new_blocks_simple. We > only need get blkoff in first group with goal and set blkoff to 0 for > the rest groups. > > Signed-off-by: Kemeng Shi <shikemeng@...weicloud.com> While this patch is OK as a simplification, there's a bigger problem with ext4_mb_new_blocks_simple(), and that is that we start looking for free blocks starting at the goal block, and then if we can't any free blocks by the time we get to the last block group.... we stop, and return ENOSPC. This function is only used in the fast commit replay path, but for a very full file system, this could cause a fast commit replay to fail unnecesarily. What we need to do is to try wrapping back to the beginning of the file system, and stop when we hit the original group (of the goal block) minus one. We can fix this up in a separate patch, since this doesn't make things any worse, but essentially we need to do something like this: maxgroups = ext4_get_groups_count(sb); for (g = 0; g < maxgroups ; g++) { ... group++; if (group > maxgroups) group = 0; } - Ted
Powered by blists - more mailing lists