Date: Mon, 16 Sep 2019 11:19:02 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Peter Jeremy <peter@rulingia.com> Cc: freebsd-current@FreeBSD.org, freebsd-arm@FreeBSD.org, rmacklem@freebsd.org Subject: Re: "Sleeping with non-sleepable lock" in NFS on recent -current Message-ID: <20190916081902.GU2559@kib.kiev.ua> In-Reply-To: <20190916074428.GF97181@server.rulingia.com> References: <20190916061205.GE97181@server.rulingia.com> <20190916063252.GS2559@kib.kiev.ua> <20190916074428.GF97181@server.rulingia.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, Sep 16, 2019 at 05:44:28PM +1000, Peter Jeremy wrote:
> On 2019-Sep-16 09:32:52 +0300, Konstantin Belousov <kostikbel@gmail.com> wrote:
> >On Mon, Sep 16, 2019 at 04:12:05PM +1000, Peter Jeremy wrote:
> >> I'm consistently seeing panics in the NFS code on recent -current on aarm64.
> >> The panics are one of the following two:
> >> Sleeping on "vmopar" with the following non-sleepable locks held:
> >> exclusive sleep mutex NEWNFSnode lock (NEWNFSnode lock) r = 0 (0xfffffd0078b346f0) locked @ /usr/src/sys/fs/nfsclient/nfs_clport.c:432
> >>
> >> Sleeping thread (tid 100077, pid 35) owns a non-sleepable lock
> >>
> >> Both panics have nearly identical backtraces (see below). I'm running
> >> diskless on a Rock64 with both filesystem and swap over NFS. The panics
> >> can be fairly reliably triggered by any of:
> >> * "make -j4 buildworld"
> >> * linking the kernel (as part of buildkernel)
> >> * "make installworld"
> >>
> >> Has anyone else seen this?
> ...
>
> >Weird since this should have been fixed long time ago. Anyway, please
> >try the following, it should fix the rest of cases.
> >
> >diff --git a/sys/fs/nfsclient/nfs_clport.c b/sys/fs/nfsclient/nfs_clport.c
> ...
> >@@ -540,7 +541,7 @@ nfscl_loadattrcache(struct vnode **vpp, struct nfsvattr *nap, void *nvaper,
> > } else {
> > np->n_size = vap->va_size;
> > np->n_flag |= NSIZECHANGED;
> >- vnode_pager_setsize(vp, np->n_size);
> >+ setnsize = 1;
>
> Should this else block include a "nsize = np->n_size;"? Without it,
> nsize will remain set to 0, which looks wrong.
Yes, you are right.
diff --git a/sys/fs/nfsclient/nfs_clport.c b/sys/fs/nfsclient/nfs_clport.c
index 471e029a8b5..63ea4736707 100644
--- a/sys/fs/nfsclient/nfs_clport.c
+++ b/sys/fs/nfsclient/nfs_clport.c
@@ -511,10 +511,10 @@ nfscl_loadattrcache(struct vnode **vpp, struct nfsvattr *nap, void *nvaper,
* zero np->n_attrstamp to indicate that
* the attributes are stale.
*/
- vap->va_size = np->n_size;
+ nsize = vap->va_size = np->n_size;
+ setnsize = 1;
np->n_attrstamp = 0;
KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
- vnode_pager_setsize(vp, np->n_size);
} else if (np->n_flag & NMODIFIED) {
/*
* We've modified the file: Use the larger
@@ -526,7 +526,8 @@ nfscl_loadattrcache(struct vnode **vpp, struct nfsvattr *nap, void *nvaper,
np->n_size = vap->va_size;
np->n_flag |= NSIZECHANGED;
}
- vnode_pager_setsize(vp, np->n_size);
+ nsize = np->n_size;
+ setnsize = 1;
} else if (vap->va_size < np->n_size) {
/*
* When shrinking the size, the call to
@@ -538,9 +539,9 @@ nfscl_loadattrcache(struct vnode **vpp, struct nfsvattr *nap, void *nvaper,
np->n_flag |= NSIZECHANGED;
setnsize = 1;
} else {
- np->n_size = vap->va_size;
+ nsize = np->n_size = vap->va_size;
np->n_flag |= NSIZECHANGED;
- vnode_pager_setsize(vp, np->n_size);
+ setnsize = 1;
}
} else {
np->n_size = vap->va_size;
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20190916081902.GU2559>