d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Avoid a data-consistency race between write() and mmap()

Browse Source 
by ensuring that newly allocated blocks are zerod.  The
race can occur even in the case where the write covers
the entire block.

Reported by: Sven Berkvens <sven@berkvens.net>, Marc Olzheim <zlo@zlo.nu>
This commit is contained in:
Matthew Dillon 2000-12-17 23:57:05 +00:00
parent 6fce744ee3
commit 6ddaf0f45e
Notes: svn2git 2020-12-20 02:59:44 +00:00 
svn path=/head/; revision=70131
3 changed files with 27 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
sys/gnu/ext2fs/ext2_readwrite.c
View File

@ -238,10 +238,19 @@ WRITE(ap)
if (uio->uio_offset + xfersize > ip->i_size)
vnode_pager_setsize(vp, uio->uio_offset + xfersize);
/*
* Avoid a data-consistency race between write() and mmap()
* by ensuring that newly allocated blocks are zerod. The
* race can occur even in the case where the write covers
* the entire block.
*/
flags |= B_CLRBUF;
#if 0
if (fs->s_frag_size > xfersize)
flags |= B_CLRBUF;
else
flags &= ~B_CLRBUF;
#endif
error = ext2_balloc(ip,
lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);

9
sys/gnu/fs/ext2fs/ext2_readwrite.c
View File

@ -238,10 +238,19 @@ WRITE(ap)
if (uio->uio_offset + xfersize > ip->i_size)
vnode_pager_setsize(vp, uio->uio_offset + xfersize);
/*
* Avoid a data-consistency race between write() and mmap()
* by ensuring that newly allocated blocks are zerod. The
* race can occur even in the case where the write covers
* the entire block.
*/
flags |= B_CLRBUF;
#if 0
if (fs->s_frag_size > xfersize)
flags |= B_CLRBUF;
else
flags &= ~B_CLRBUF;
#endif
error = ext2_balloc(ip,
lbn, blkoffset + xfersize, ap->a_cred, &bp, flags);

9
sys/ufs/ufs/ufs_readwrite.c
View File

@ -468,10 +468,19 @@ WRITE(ap)
if (uio->uio_offset + xfersize > ip->i_size)
vnode_pager_setsize(vp, uio->uio_offset + xfersize);
/*
* Avoid a data-consistency race between write() and mmap()
* by ensuring that newly allocated blocks are zerod. The
* race can occur even in the case where the write covers
* the entire block.
*/
flags |= B_CLRBUF;
#if 0
if (fs->fs_bsize > xfersize)
flags |= B_CLRBUF;
else
flags &= ~B_CLRBUF;
#endif
/* XXX is uio->uio_offset the right thing here? */
error = VOP_BALLOC(vp, uio->uio_offset, xfersize,
ap->a_cred, flags, &bp);