NetBSD-Bugs archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
RE: kern/59452: Move NFS wcc data messages to debug or control them via sysctl.
Hi All,
Any change to apply such patch?
It's moving these errors messages into DEBUG mode.
==cut==
--- /root/nfs_clntsubs.c 2025-10-22 21:00:13.919594789 +0200
+++ nfs_clntsubs.c 2025-10-22 20:13:06.624486932 +0200
@@ -393,15 +393,20 @@
time_t now = time_second;
const struct timespec *omtime = &np->n_vattr->va_mtime;
const struct timespec *octime = &np->n_vattr->va_ctime;
+#ifdef DEBUG
const char *reason = NULL; /* XXX: gcc */
-
+#endif
if (timespeccmp(omtime, mtime, <=)) {
+#ifdef DEBUG
reason = "mtime";
+#endif
error = EINVAL;
}
if (vp->v_type == VDIR && timespeccmp(octime, ctime, <=)) {
+#ifdef DEBUG
reason = "ctime";
+#endif
error = EINVAL;
}
@@ -423,6 +428,7 @@
*/
mutex_enter(&nmp->nm_lock);
+#ifdef DEBUG
if (!NFS_WCCKLUDGE(nmp, now)) {
printf("%s: inaccurate wcc data (%s) detected,"
" disabling wcc"
@@ -439,6 +445,7 @@
(unsigned int)mtime->tv_sec,
(unsigned int)mtime->tv_nsec);
}
+#endif
nmp->nm_iflag |= NFSMNT_WCCKLUDGE;
nmp->nm_wcckludgetime = now;
mutex_exit(&nmp->nm_lock);
@@ -447,8 +454,10 @@
} else if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) {
mutex_enter(&nmp->nm_lock);
if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) {
+#ifdef DEBUG
printf("%s: re-enabling wcc\n",
vp->v_mount->mnt_stat.f_mntfromname);
+#endif
nmp->nm_iflag &= ~NFSMNT_WCCKLUDGE;
}
mutex_exit(&nmp->nm_lock);
==cut==
--
Marcin Gondek / Drixter
http://fido.e-utp.net/
AS56662
-----Original Message-----
From: netbsd-bugs-owner%NetBSD.org@localhost <netbsd-bugs-owner%NetBSD.org@localhost> On Behalf Of Marcin Gondek via gnats
Sent: Wednesday, October 22, 2025 9:10 PM
To: kern-bug-people%netbsd.org@localhost; gnats-admin%netbsd.org@localhost; netbsd-bugs%netbsd.org@localhost; Marcin Gondek <drixter%e-utp.net@localhost>
Subject: RE: kern/59452: Move NFS wcc data messages to debug or control them via sysctl.
The following reply was made to PR kern/59452; it has been noted by GNATS.
From: Marcin Gondek <drixter%e-utp.net@localhost>
To: "gnats-bugs%netbsd.org@localhost" <gnats-bugs%netbsd.org@localhost>,
"kern-bug-people%netbsd.org@localhost" <kern-bug-people%netbsd.org@localhost>,
"gnats-admin%netbsd.org@localhost" <gnats-admin%netbsd.org@localhost>, "netbsd-bugs%netbsd.org@localhost"
<netbsd-bugs%netbsd.org@localhost>
Cc:
Subject: RE: kern/59452: Move NFS wcc data messages to debug or control them via sysctl.
Date: Wed, 22 Oct 2025 19:08:42 +0000
Hi All,
I've made a patch for myself to disable it via ifdef, probably this is not = proper way, but my system is getting spam on high load over NFS.
Now observing that is better or not, as far I'm not C developer I hope that=
there is no other side effect.
The good thing is that kernel compile correctly. First my approach was only=
theory now time for practice ;-)
=3D=3Dcut=3D=3D
--- nfs_clntsubs.c 2025-10-22 21:00:13.919594789 +0200
+++ nfs_clntsubs.c-IFDEF 2025-10-22 19:38:56.907354570 +0200
@@ -393,15 +393,20 @@
time_t now =3D time_second;
const struct timespec *omtime =3D &np->n_vattr->va_mtime;
const struct timespec *octime =3D &np->n_vattr->va_ctime; +#ifdef DEBUG
const char *reason =3D NULL; /* XXX: gcc */
-
+#endif
if (timespeccmp(omtime, mtime, <=3D)) { +#ifdef DEBUG
reason =3D "mtime"; +#endif
error =3D EINVAL;
}
if (vp->v_type =3D=3D VDIR && timespeccmp(octime, ctime, <=
=3D)) {
+#ifdef DEBUG
reason =3D "ctime"; +#endif
error =3D EINVAL;
}
@@ -423,6 +428,7 @@
*/
mutex_enter(&nmp->nm_lock); +#ifdef DEBUG
if (!NFS_WCCKLUDGE(nmp, now)) {
printf("%s: inaccurate wcc data (%s) detect= ed,"
" disabling wcc"
@@ -439,6 +445,7 @@
(unsigned int)mtime->tv_sec,
(unsigned int)mtime->tv_nsec);
}
+#endif
nmp->nm_iflag |=3D NFSMNT_WCCKLUDGE;
nmp->nm_wcckludgetime =3D now;
mutex_exit(&nmp->nm_lock); @@ -447,8 +454,10 @@
} else if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) {
mutex_enter(&nmp->nm_lock);
if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) { +#ifdef DEBUG
printf("%s: re-enabling wcc\n",
vp->v_mount->mnt_stat.f_mntfromname);
+#endif
nmp->nm_iflag &=3D ~NFSMNT_WCCKLUDGE;
}
mutex_exit(&nmp->nm_lock); =3D=3Dcut=3D=3D
--
Marcin Gondek / Drixter
http://fido.e-utp.net/
AS56662
-----Original Message-----
From: netbsd-bugs-owner%NetBSD.org@localhost <netbsd-bugs-owner%NetBSD.org@localhost> On Behalf=
Of Marcin Gondek
Sent: Thursday, July 24, 2025 10:17 AM
To: gnats-bugs%netbsd.org@localhost; kern-bug-people%netbsd.org@localhost; gnats-admin@netbsd.o= rg; netbsd-bugs%netbsd.org@localhost
Subject: RE: kern/59452: Move NFS wcc data messages to debug or control the= m via sysctl.
Hi All,
Is there any chance that someone will change it?
I know that is rare case, but for my boxes I'm getting spam in dmesg on eve= ry high load ;-/ With noac option there is a less spam but performance is s= ignificantly degraded.
Thanks,
--
Marcin Gondek / Drixter
http://fido.e-utp.net/
AS56662
-----Original Message-----
From: Marcin Gondek
Sent: Monday, June 9, 2025 12:48 PM
To: Marcin Gondek <drixter%e-utp.net@localhost>; gnats-bugs%netbsd.org@localhost; kern-bug-peop= le%netbsd.org@localhost; gnats-admin%netbsd.org@localhost; netbsd-bugs%netbsd.org@localhost
Subject: RE: kern/59452: Move NFS wcc data messages to debug or control the= m via sysctl.
Hi.
[ 190.659022] diskstation6.e-utp.net:/volume1/homes/drixter: inaccurate w=
cc data (ctime) detected, disabling wcc (ctime 1748779290.941608597 1748779= 290.941608597, mtime 1748779290.941608597 1748779290.941608597) [ 14577.965578] diskstation6.e-utp.net:/volume1/Applications/NetBSD: inaccu= rate wcc data (ctime) detected, disabling wcc (ctime 1748793678.251985388 1= 748793678.251985388, mtime 1748793678.251985388 1748793678.251985388) [ 398= 26.666864] diskstation6.e-utp.net:/volume1/Applications/html: inaccurate wc= c data (ctime) detected, disabling wcc (ctime 1748818926.951985610 17488189= 26.951985610, mtime 1748818926.951985610 1748818926.951985610) [ 103278.442= 441] diskstation6.e-utp.net:/volume1/Applications/NetBSD: re-enabling wcc [=
104091.099550] diskstation6.e-utp.net:/volume1/Applications/NetBSD: inaccu= rate wcc data (ctime) detected, disabling wcc (ctime 1748883191.385396294 1= 748883191.385396294, mtime 1748883191.385396294 1748883191.385396294) [ 112= 832.426073] diskstation6.e-utp.net:/volume1/homes/drixter: re-enabling wcc = [ 112853.086255] diskstation6.e-utp.net:/volume1/homes/drixter: inaccurate = wcc data (ctime) detected, disabling wcc (ctime 1748891953.379970261 174889= 1953.379970261, mtime 1748891953.379970261 1748891953.379970261) [ 530900.5= 89197] diskstation6.e-utp.net:/volume1/Applications/NetBSD: re-enabling wcc=
[ 530982.569922] diskstation6.e-utp.net:/volume1/Applications/NetBSD: inac= curate wcc data (ctime) detected, disabling wcc (ctime 1749310082.858985530=
1749310082.858985530, mtime 1749310082.858985530 1749310082.858985530)
The logs messages in kernel, still appears, even with noac option, the side= -effect is that NFS is quite slow right now when processing multiple small = files.
I suggest maybe to move forward with my idea to control it via DEBUG and./o= r sysctl variable if possible.
Thanks,
--
Marcin Gondek / Drixter
http://fido.e-utp.net/
AS56662
-----Original Message-----
From: netbsd-bugs-owner%NetBSD.org@localhost <netbsd-bugs-owner%NetBSD.org@localhost> On Behalf=
Of Marcin Gondek
Sent: Sunday, June 1, 2025 2:07 PM
To: gnats-bugs%netbsd.org@localhost; kern-bug-people%netbsd.org@localhost; gnats-admin@netbsd.o= rg; netbsd-bugs%netbsd.org@localhost
Subject: RE: kern/59452: Move NFS wcc data messages to debug or control the= m via sysctl.
Hi Michael
I've setup now as you suggested, It's the same, box restarted to be sure:
[ 190.659022] diskstation6.e-utp.net:/volume1/homes/drixter: inaccurate w=
cc data (ctime) detected, disabling wcc (ctime 1748779290.941608597 1748779= 290.941608597, mtime 1748779290.941608597 1748779290.941608597)
devil# mount -vv | grep drixter | grep noac diskstation6.e-utp.net:/volume1= /homes/drixter on /home/users/drixter type nfs (automounted, fsid: 0xb0e/0x= 70b, reads: sync 0 async 0, writes: sync 0 async 0, [nfs: addr=3D2001:67c:2= 1ec:ffee:18f2:a77c:dc08:aef6, port=3D2049, addrlen=3D28, sotype=3D1, proto= =3D0, fhsize=3D0, flags=3D0x88247<soft,wsize,rsize,intr,nfsv3,resvport,noac=
>, wsize=3D8192, rsize=3D8192, readdirsize=3D4096, timeo=3D300, retrans=3D1= 0, maxgrouplist=3D16, readahead=3D2, leaseterm=3D0, deadthresh=3D9]) devil#
devil# cat /etc/auto_homes
* -rw,noac,soft,intr,rsize=3D8192,wsize=3D8192 diskstation6.e-utp.net:/volu= me1/homes/&
noac is included.
Thanks,
--
Marcin Gondek / Drixter
http://fido.e-utp.net/
AS56662
-----Original Message-----
From: netbsd-bugs-owner%NetBSD.org@localhost <netbsd-bugs-owner%NetBSD.org@localhost> On Behalf=
Of Michael van Elst via gnats
Sent: Sunday, June 1, 2025 1:45 PM
To: kern-bug-people%netbsd.org@localhost; gnats-admin%netbsd.org@localhost; netbsd-bugs@netbsd.= org; Marcin Gondek <drixter%e-utp.net@localhost>
Subject: Re: kern/59452: Move NFS wcc data messages to debug or control the= m via sysctl.
The following reply was made to PR kern/59452; it has been noted by GNATS.
From: mlelstv%serpens.de@localhost (Michael van Elst)
To: gnats-bugs%netbsd.org@localhost
Cc:=20
Subject: Re: kern/59452: Move NFS wcc data messages to debug or control the= m via sysctl.
Date: Sun, 1 Jun 2025 11:41:06 -0000 (UTC)
drixter%e-utp.net@localhost writes:
=20
>I know that such messages should not generally appear if the NFS server i= s working properly. But unfortunately it is not repairable everywhere -> Sy= nology NFS, hence the question about the possibility of moving these messag= es to the DEBUG variant or maybe make it controlable via sysctl.
=20
>[ 514378.384585] diskstation6.e-utp.net:/volume1/Applications/NetBSD: re-= enabling wcc >[ 514557.066122] diskstation6.e-utp.net:/volume1/Application=
s/NetBSD: inaccurate wcc data (ctime) detected, disabling wcc (ctime 174837=
4671.270579777 1748374671.270579777, mtime 1748374671.270579777 1748374671.=
270579777)
=20
Most of the messages appear to be "re-enabling" + "disabling wcc" again wi= thin a few seconds.
=20
What happens when you just disable the attribute cache when mounting with = 'mount -o noac' ? It's hardly used anyway.
=20
Home |
Main Index |
Thread Index |
Old Index