linux-ext4 - Re: [PATCH 0/5] mballoc: trivial code cleanup

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 linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <AANLkTimT3i1oOSh02oTtvN_OmHcYwhNOABLV4cT7VV5p@mail.gmail.com>
Date:	Sun, 6 Feb 2011 10:43:58 +0200
From:	Amir Goldstein <amir73il@...il.com>
To:	i@...y.li
Cc:	linux-ext4@...r.kernel.org, Alex Tomas <alex@...sterfs.com>,
	Theodore Tso <tytso@...gle.com>
Subject: Re: [PATCH 0/5] mballoc: trivial code cleanup

Hi Coly,

Forgive me for hijacking your thread.
I cannot comment on the cleanup, as I am new to mballoc code,
but I was wondering if you could offer me a piece of advise.

In the current implementation (prototype) of ext4 snapshots,
I have an outstanding circular lockdep warning to fix:
inode->i_data_sem => grp->alloc_sem => snapshot->i_data_sem/1 => grp->alloc_sem

Snapshot block allocation can be nested inside another inode block allocation,
because COW of bitmap block is triggered on get_write_access() inside
ext4_mb_mark_diskspace_used().
i_data_sem nested lockdep is handled by calling
down_read/write_nested() in ext4_map_blocks():
https://github.com/amir73il/ext4-snapshots/blob/ext4-snapshots/fs/ext4/inode.c#L1540

In ext3/next3, handling nested i_truncate_mutex lockdep was enough,
but in ext4, I need to take care of alloc_sem and lg_mutex as well.
I "handled" lg_mutex by never setting EXT4_MB_HINT_DATA on snapshot files.
When looking at alloc_sem, I realized that it is only needed to avoid
race with adjacent group buddy initialization.
Snapshots feature, however, requires that block_size == page_size, so
it seems that alloc_sem
can be avoided altogether.

My questions are:

1. Am I missing something in my analysis?
2. Wouldn't it make sense to bypass alloc_sem at all times if
block_size == page_size
(or block_size*2 >= page_size), regardless of the snapshots feature?
After all, a group buddy is always initialized before calling
allocation routines.

Thanks for your time,
Amir.

On Sun, Feb 6, 2011 at 6:53 AM, Coly Li <i@...y.li> wrote:
> This patch set contains several trivial code cleanup to mballoc code.
> A basic testing is done with Linux 2.6.38-rc3+ (commit 44f2c5c8).
>
> Signed-off-by: Coly Li <bosong.ly@...bao.com>
> Cc: Alex Tomas <alex@...sterfs.com>
> Cc: Theodore Tso <tytso@...gle.com>
> ---
>  fs/ext4/mballoc.c |   15 ++++++++++-----
>  fs/ext4/mballoc.h |    2 +-
>  2 files changed, 11 insertions(+), 6 deletions(-)
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to majordomo@...r.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html