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

nfscl: Check for mmap(2)'d file before doing direct output

Browse Source 
Commit 867c27c23a modified the NFS client so that
it does IO_APPEND writes directly to the NFS server,
bypassing the buffer cache.  However, this could result
in stale data in client pages when the file is mmap(2)'d.
As such, the NFS client needs to call vm_object_is_active()
to check if the file is mmap(2)'d and only do direct
output if the file is not mmap(2)'d.

This patch adds this check.

Although a simple patch, I have given it a long MFC,
since the related commit 867c27c23a made a significant
semantics change and, as such, has a long MFC.

MFC after:	3 months
This commit is contained in:
Rick Macklem 2021-12-20 13:08:51 -08:00
parent d69b9cc26d
commit b70042adfe
1 changed files with 13 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
sys/fs/nfsclient/nfs_clbio.c
View File

@ -1001,12 +1001,24 @@ ncl_write(struct vop_write_args *ap)
if (uio->uio_resid == 0)
return (0);
/*
* If the file in not mmap()'d, do IO_APPEND writing via a
* synchronous direct write. This can result in a significant
* performance improvement.
* If the file is mmap()'d, this cannot be done, since there
* is no way to maintain consistency between the file on the
* NFS server and the file's mmap()'d pages.
*/
NFSLOCKNODE(np);
if (vp->v_type == VREG && ((newnfs_directio_enable && (ioflag &
IO_DIRECT)) || (ioflag & IO_APPEND))) {
IO_DIRECT)) || ((ioflag & IO_APPEND) &&
(vp->v_object == NULL || !vm_object_is_active(vp->v_object))))) {
NFSUNLOCKNODE(np);
if ((ioflag & IO_APPEND) != 0)
ioflag |= IO_SYNC;
return nfs_directio_write(vp, uio, cred, ioflag);
}
NFSUNLOCKNODE(np);
/*
* Maybe this should be above the vnode op call, but so long as