system/freebsd-src
system/freebsd-src
1
0
Fork 0
mirror of https://github.com/freebsd/freebsd-src synced 2024-07-24 03:37:16 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

nfscl: Do not do readahead for directories

Browse source 
For a very long time, the NFS client has done readahead for
directory blocks.  Unlike data blocks, the readahead cannot
begin until the Readdir RPC reply has been received, since
the directory offset cookie in that Readdir RPC reply is needed.
As such, the readahead is serialized and does not seem to
provide any real benefit.

Recent testing/benchmarking shows that removing this
readahead code for Readdir does not have a negative impact
on performance.

Therefore, this patch deletes the readahead code for Readdir,
which simplifies the code and may make future changes simpler.

MFC after:	1 month
This commit is contained in:
Rick Macklem 2024-05-09 18:33:13 -07:00
parent f77b5b295d
commit fbe965591f
1 changed files with 1 additions and 30 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

31
sys/fs/nfsclient/nfs_clbio.c
View file

@ -679,36 +679,6 @@ ncl_bioread(struct vnode *vp, struct uio *uio, int ioflag, struct ucred *cred)
goto out;
}
/*
* If not eof and read aheads are enabled, start one.
* (You need the current block first, so that you have the
* directory offset cookie of the next block.)
*/
NFSLOCKNODE(np);
if (nmp->nm_readahead > 0 && ncl_bioread_dora(vp) &&
(bp->b_flags & B_INVAL) == 0 &&
(np->n_direofoffset == 0 ||
(lbn + 1) * NFS_DIRBLKSIZ < np->n_direofoffset) &&
incore(&vp->v_bufobj, lbn + 1) == NULL) {
NFSUNLOCKNODE(np);
rabp = nfs_getcacheblk(vp, lbn + 1, NFS_DIRBLKSIZ, td);
if (rabp) {
if ((rabp->b_flags & (B_CACHE|B_DELWRI)) == 0) {
rabp->b_flags |= B_ASYNC;
rabp->b_iocmd = BIO_READ;
vfs_busy_pages(rabp, 0);
if (ncl_asyncio(nmp, rabp, cred, td)) {
rabp->b_flags |= B_INVAL;
rabp->b_ioflags |= BIO_ERROR;
vfs_unbusy_pages(rabp);
brelse(rabp);
}
} else {
brelse(rabp);
}
}
NFSLOCKNODE(np);
}
/*
* Unlike VREG files, whos buffer size ( bp->b_bcount ) is
* chopped for the EOF condition, we cannot tell how large
@ -721,6 +691,7 @@ ncl_bioread(struct vnode *vp, struct uio *uio, int ioflag, struct ucred *cred)
* in np->n_direofoffset and chop it off as an extra step
* right here.
*/
NFSLOCKNODE(np);
n = lmin(uio->uio_resid, NFS_DIRBLKSIZ - bp->b_resid - on);
if (np->n_direofoffset && n > np->n_direofoffset - uio->uio_offset)
n = np->n_direofoffset - uio->uio_offset;