diff options
| author | rwatson <rwatson@FreeBSD.org> | 2003-06-11 03:32:42 +0000 |
|---|---|---|
| committer | rwatson <rwatson@FreeBSD.org> | 2003-06-11 03:32:42 +0000 |
| commit | decffe61325704d22caa6b0de43e88b38ce47201 (patch) | |
| tree | de9192844197d3bf4ebe26487f94a773a4eeef77 /sys/nfsclient | |
| parent | 52433d15d4ea215b1cf0f613c42d2d5cbfdf1b81 (diff) | |
| download | FreeBSD-src-decffe61325704d22caa6b0de43e88b38ce47201.zip FreeBSD-src-decffe61325704d22caa6b0de43e88b38ce47201.tar.gz | |
Add the comment I meant to add about not passing in PCATCH to the
tsleep(). Note the XXX.
Diffstat (limited to 'sys/nfsclient')
| -rw-r--r-- | sys/nfsclient/nfs_lock.c | 8 |
1 files changed, 8 insertions, 0 deletions
diff --git a/sys/nfsclient/nfs_lock.c b/sys/nfsclient/nfs_lock.c index ba10926..f524712 100644 --- a/sys/nfsclient/nfs_lock.c +++ b/sys/nfsclient/nfs_lock.c @@ -190,6 +190,14 @@ nfs_dolock(struct vop_advlock_args *ap) * then even a reasonably loaded system should take (at least * on a local network). XXX Probably should use a back-off * scheme. + * + * XXX: No PCATCH here since we currently have no useful + * way to signal to the userland rpc.lockd that the request + * has been aborted. Once the rpc.lockd implementation + * can handle aborts, and we report them properly, + * PCATCH can be put back. In the mean time, if we did + * permit aborting, the lock attempt would "get lost" + * and the lock would get stuck in the locked state. */ error = tsleep(p->p_nlminfo, PUSER, "lockd", 20*hz); if (error != 0) { |
