From: Jan Kara <jack@suse.cz>
To: Ingo Oeser <ioe-lkml@rameria.de>
Cc: LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH REVIEW] udf: Fix lock inversion between iprune_mutex and alloc_mutex
Date: Mon, 18 Aug 2008 14:47:59 +0200 [thread overview]
Message-ID: <20080818124758.GA11523@duck.suse.cz> (raw)
In-Reply-To: <200808142048.10966.ioe-lkml@rameria.de>
On Thu 14-08-08 20:48:10, Ingo Oeser wrote:
> Hi Jan,
>
> On Thursday 14 August 2008, Jan Kara wrote:
> > A memory allocation inside alloc_mutex must not recurse back
> > into the filesystem itself because that leads to lock inversion
> > between iprune_mutex and alloc_mutex (and thus to deadlocks - see
> > traces below). Make allocations inside alloc_mutex use GFP_NOFS
> > to avoid this problem.
>
> What about moving the allocation before taking the mutex?
> That way, we even reduce the lock contention and simplify the failure path.
Thanks for the idea and the review. Actually, I've checked and
i_alloc_mutex is needed only to update allocation statistics in the
superblock. So we can release i_alloc_mutex earlier before allocating
memory. Actually, I ended up moving the memory allocation as well - even
before the block allocation - to fix the error path cleanup. I'll post
resulting two patches in a moment.
Honza
> Signed-off-by: Ingo Oeser <ioe-lkml@rameria.de>
>
> diff --git a/fs/udf/ialloc.c b/fs/udf/ialloc.c
> index eb9cfa2..76446eb 100644
> --- a/fs/udf/ialloc.c
> +++ b/fs/udf/ialloc.c
> @@ -90,6 +90,25 @@ struct inode *udf_new_inode(struct inode *dir, int mode, int *err)
> return NULL;
> }
>
> + if (UDF_QUERY_FLAG(inode->i_sb, UDF_FLAG_USE_EXTENDED_FE)) {
> + iinfo->i_efe = 1;
> + if (UDF_VERS_USE_EXTENDED_FE > sbi->s_udfrev)
> + sbi->s_udfrev = UDF_VERS_USE_EXTENDED_FE;
> + iinfo->i_ext.i_data = kzalloc(inode->i_sb->s_blocksize -
> + sizeof(struct extendedFileEntry),
> + GFP_KERNEL);
> + } else {
> + iinfo->i_efe = 0;
> + iinfo->i_ext.i_data = kzalloc(inode->i_sb->s_blocksize -
> + sizeof(struct fileEntry),
> + GFP_KERNEL);
> + }
> + if (!iinfo->i_ext.i_data) {
> + iput(inode);
> + *err = -ENOMEM;
> + return NULL;
> + }
> +
> mutex_lock(&sbi->s_alloc_mutex);
> if (sbi->s_lvid_bh) {
> struct logicalVolIntegrityDesc *lvid =
> @@ -129,25 +148,6 @@ struct inode *udf_new_inode(struct inode *dir, int mode, int *err)
> iinfo->i_lenEAttr = 0;
> iinfo->i_lenAlloc = 0;
> iinfo->i_use = 0;
> - if (UDF_QUERY_FLAG(inode->i_sb, UDF_FLAG_USE_EXTENDED_FE)) {
> - iinfo->i_efe = 1;
> - if (UDF_VERS_USE_EXTENDED_FE > sbi->s_udfrev)
> - sbi->s_udfrev = UDF_VERS_USE_EXTENDED_FE;
> - iinfo->i_ext.i_data = kzalloc(inode->i_sb->s_blocksize -
> - sizeof(struct extendedFileEntry),
> - GFP_KERNEL);
> - } else {
> - iinfo->i_efe = 0;
> - iinfo->i_ext.i_data = kzalloc(inode->i_sb->s_blocksize -
> - sizeof(struct fileEntry),
> - GFP_KERNEL);
> - }
> - if (!iinfo->i_ext.i_data) {
> - iput(inode);
> - *err = -ENOMEM;
> - mutex_unlock(&sbi->s_alloc_mutex);
> - return NULL;
> - }
> if (UDF_QUERY_FLAG(inode->i_sb, UDF_FLAG_USE_AD_IN_ICB))
> iinfo->i_alloc_type = ICBTAG_FLAG_AD_IN_ICB;
> else if (UDF_QUERY_FLAG(inode->i_sb, UDF_FLAG_USE_SHORT_AD))
--
Jan Kara <jack@suse.cz>
SUSE Labs, CR
prev parent reply other threads:[~2008-08-18 12:48 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-08-14 17:49 [PATCH REVIEW] udf: Fix lock inversion between iprune_mutex and alloc_mutex Jan Kara
2008-08-14 18:48 ` Ingo Oeser
2008-08-18 12:47 ` Jan Kara [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20080818124758.GA11523@duck.suse.cz \
--to=jack@suse.cz \
--cc=ioe-lkml@rameria.de \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).