| Message ID | 1394143352-26095-1-git-send-email-rees@umich.edu (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, 6 Mar 2014 17:02:32 -0500 Jim Rees <rees@umich.edu> wrote: > Signed-off-by: Jim Rees <rees@umich.edu> > --- > utils/mount/nfs.man | 53 ++++++----------------------------------------------- > 1 file changed, 6 insertions(+), 47 deletions(-) > > diff --git a/utils/mount/nfs.man b/utils/mount/nfs.man > index ef09a31..2ab369c 100644 > --- a/utils/mount/nfs.man > +++ b/utils/mount/nfs.man > @@ -125,6 +125,12 @@ option may mitigate some of the risks of using the > .B soft > option. > .TP 1.5i > +.BR intr " / " nointr > +This option is provided for backward compatibility. > +It is ignored after kernel 2.6.25. > +System calls return EINTR if an in-progress NFS operation is interrupted by > +a signal. That isn't correct. "A process that is waiting for a reply for an NFS server can be killed by any signal which would normally kill the process. If a signal would not normally kill the process (i.e. it is caught or ignored) then that signal will not abort and NFS request". This is really a fairly horrible semantic. It makes perfect sense from an internal implementation perspective and provides good backwards compatibility and cross-compatibility with other filesystems (where 'read' and 'write' never ever return EINTR), but it is very hard to document clearly. NeilBrown > +.TP 1.5i > .BI timeo= n > The time in deciseconds (tenths of a second) the NFS client waits for a > response before it retries an NFS request. > @@ -668,30 +674,6 @@ Using the > option is also required when mounting exports on NFS servers > that do not support the NLM protocol. > .TP 1.5i > -.BR intr " / " nointr > -Selects whether to allow signals to interrupt file operations > -on this mount point. If neither option > -is specified (or if > -.B nointr > -is specified), > -signals do not interrupt NFS file operations. If > -.B intr > -is specified, system calls return EINTR if an in-progress NFS operation is interrupted by > -a signal. > -.IP > -Using the > -.B intr > -option is preferred to using the > -.B soft > -option because it is significantly less likely to result in data corruption. > -.IP > -The > -.BR intr " / " nointr > -mount option is deprecated after kernel 2.6.25. > -Only SIGKILL can interrupt a pending NFS operation on these kernels, > -and if specified, this mount option is ignored to provide backwards > -compatibility with older kernels. > -.TP 1.5i > .BR cto " / " nocto > Selects whether to use close-to-open cache coherence semantics. > If neither option is specified (or if > @@ -807,29 +789,6 @@ The mount request fails if the server's rpcbind service is not available, > the server's NFS service is not registered with its rpcbind service, > or the server's NFS service is not available on the advertised port. > .TP 1.5i > -.BR intr " / " nointr > -Selects whether to allow signals to interrupt file operations > -on this mount point. If neither option is specified (or if > -.B intr > -is specified), system calls return EINTR if an in-progress NFS operation > -is interrupted by a signal. If > -.B nointr > -is specified, signals do not > -interrupt NFS operations. > -.IP > -Using the > -.B intr > -option is preferred to using the > -.B soft > -option because it is significantly less likely to result in data corruption. > -.IP > -The > -.BR intr " / " nointr > -mount option is deprecated after kernel 2.6.25. > -Only SIGKILL can interrupt a pending NFS operation on these kernels, > -and if specified, this mount option is ignored to provide backwards > -compatibility with older kernels. > -.TP 1.5i > .BR cto " / " nocto > Selects whether to use close-to-open cache coherence semantics > for NFS directories on this mount point.
On Mar 7, 2014, at 2:59, NeilBrown <neilb@suse.de> wrote: > On Thu, 6 Mar 2014 17:02:32 -0500 Jim Rees <rees@umich.edu> wrote: > >> Signed-off-by: Jim Rees <rees@umich.edu> >> --- >> utils/mount/nfs.man | 53 ++++++----------------------------------------------- >> 1 file changed, 6 insertions(+), 47 deletions(-) >> >> diff --git a/utils/mount/nfs.man b/utils/mount/nfs.man >> index ef09a31..2ab369c 100644 >> --- a/utils/mount/nfs.man >> +++ b/utils/mount/nfs.man >> @@ -125,6 +125,12 @@ option may mitigate some of the risks of using the >> .B soft >> option. >> .TP 1.5i >> +.BR intr " / " nointr >> +This option is provided for backward compatibility. >> +It is ignored after kernel 2.6.25. >> +System calls return EINTR if an in-progress NFS operation is interrupted by >> +a signal. > > That isn't correct. > "A process that is waiting for a reply for an NFS server can be killed > by any signal which would normally kill the process. If a signal would not > normally kill the process (i.e. it is caught or ignored) then that signal > will not abort and NFS request". > > This is really a fairly horrible semantic. It makes perfect sense from an > internal implementation perspective and provides good backwards compatibility > and cross-compatibility with other filesystems (where 'read' and 'write' > never ever return EINTR), but it is very hard to document clearly. > Isn’t the documentation in the signal(7) manpage sufficient? We could simply refer to that…
On Fri, 7 Mar 2014 08:13:00 -0500 Trond Myklebust <trond.myklebust@primarydata.com> wrote: > > On Mar 7, 2014, at 2:59, NeilBrown <neilb@suse.de> wrote: > > > On Thu, 6 Mar 2014 17:02:32 -0500 Jim Rees <rees@umich.edu> wrote: > > > >> Signed-off-by: Jim Rees <rees@umich.edu> > >> --- > >> utils/mount/nfs.man | 53 ++++++----------------------------------------------- > >> 1 file changed, 6 insertions(+), 47 deletions(-) > >> > >> diff --git a/utils/mount/nfs.man b/utils/mount/nfs.man > >> index ef09a31..2ab369c 100644 > >> --- a/utils/mount/nfs.man > >> +++ b/utils/mount/nfs.man > >> @@ -125,6 +125,12 @@ option may mitigate some of the risks of using the > >> .B soft > >> option. > >> .TP 1.5i > >> +.BR intr " / " nointr > >> +This option is provided for backward compatibility. > >> +It is ignored after kernel 2.6.25. > >> +System calls return EINTR if an in-progress NFS operation is interrupted by > >> +a signal. > > > > That isn't correct. > > "A process that is waiting for a reply for an NFS server can be killed > > by any signal which would normally kill the process. If a signal would not > > normally kill the process (i.e. it is caught or ignored) then that signal > > will not abort and NFS request". > > > > This is really a fairly horrible semantic. It makes perfect sense from an > > internal implementation perspective and provides good backwards compatibility > > and cross-compatibility with other filesystems (where 'read' and 'write' > > never ever return EINTR), but it is very hard to document clearly. > > > Isn’t the documentation in the signal(7) manpage sufficient? We could simply refer to that… > I would certainly be good if we could delegation the documentation. But I couldn't find anything in signal(7) that seemed particularly relevant. Which bit were you thinking of? Thanks, NeilBrown
On Mar 7, 2014, at 18:42, NeilBrown <neilb@suse.de> wrote: > On Fri, 7 Mar 2014 08:13:00 -0500 Trond Myklebust > <trond.myklebust@primarydata.com> wrote: > >> >> On Mar 7, 2014, at 2:59, NeilBrown <neilb@suse.de> wrote: >> >>> On Thu, 6 Mar 2014 17:02:32 -0500 Jim Rees <rees@umich.edu> wrote: >>> >>>> Signed-off-by: Jim Rees <rees@umich.edu> >>>> --- >>>> utils/mount/nfs.man | 53 ++++++----------------------------------------------- >>>> 1 file changed, 6 insertions(+), 47 deletions(-) >>>> >>>> diff --git a/utils/mount/nfs.man b/utils/mount/nfs.man >>>> index ef09a31..2ab369c 100644 >>>> --- a/utils/mount/nfs.man >>>> +++ b/utils/mount/nfs.man >>>> @@ -125,6 +125,12 @@ option may mitigate some of the risks of using the >>>> .B soft >>>> option. >>>> .TP 1.5i >>>> +.BR intr " / " nointr >>>> +This option is provided for backward compatibility. >>>> +It is ignored after kernel 2.6.25. >>>> +System calls return EINTR if an in-progress NFS operation is interrupted by >>>> +a signal. >>> >>> That isn't correct. >>> "A process that is waiting for a reply for an NFS server can be killed >>> by any signal which would normally kill the process. If a signal would not >>> normally kill the process (i.e. it is caught or ignored) then that signal >>> will not abort and NFS request". >>> >>> This is really a fairly horrible semantic. It makes perfect sense from an >>> internal implementation perspective and provides good backwards compatibility >>> and cross-compatibility with other filesystems (where 'read' and 'write' >>> never ever return EINTR), but it is very hard to document clearly. >>> >> Isn’t the documentation in the signal(7) manpage sufficient? We could simply refer to that… >> > > I would certainly be good if we could delegation the documentation. But I > couldn't find anything in signal(7) that seemed particularly relevant. Which > bit were you thinking of? > The table which lists the posix signals, and which indicates whether they are fatal not (i.e. what ‘action’ they cause). The same page also explains signal blocking and signal handling, which could also be helpful..
Trond Myklebust wrote: > I would certainly be good if we could delegation the documentation. But I > couldn't find anything in signal(7) that seemed particularly relevant. Which > bit were you thinking of? > The table which lists the posix signals, and which indicates whether they are fatal not (i.e. what ‘action’ they cause). The same page also explains signal blocking and signal handling, which could also be helpful.. This seems the relevant part to me, assuming nfs system calls are considered "slow": * read(2), readv(2), write(2), writev(2), and ioctl(2) calls on "slow" devices. A "slow" device is one where the I/O call may block for an indefinite time, for example, a terminal, pipe, or socket. (A disk is not a slow device according to this defini? tion.) If an I/O call on a slow device has already transferred some data by the time it is interrupted by a signal handler, then the call will return a success status (normally, the number of bytes transferred). Here's Neil's description: >>> "A process that is waiting for a reply for an NFS server can be killed >>> by any signal which would normally kill the process. If a signal would not >>> normally kill the process (i.e. it is caught or ignored) then that signal >>> will not abort and NFS request". So assuming nfs system calls honor the SA_RESTART flag, I think these behaviors are the same? -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/utils/mount/nfs.man b/utils/mount/nfs.man index ef09a31..2ab369c 100644 --- a/utils/mount/nfs.man +++ b/utils/mount/nfs.man @@ -125,6 +125,12 @@ option may mitigate some of the risks of using the .B soft option. .TP 1.5i +.BR intr " / " nointr +This option is provided for backward compatibility. +It is ignored after kernel 2.6.25. +System calls return EINTR if an in-progress NFS operation is interrupted by +a signal. +.TP 1.5i .BI timeo= n The time in deciseconds (tenths of a second) the NFS client waits for a response before it retries an NFS request. @@ -668,30 +674,6 @@ Using the option is also required when mounting exports on NFS servers that do not support the NLM protocol. .TP 1.5i -.BR intr " / " nointr -Selects whether to allow signals to interrupt file operations -on this mount point. If neither option -is specified (or if -.B nointr -is specified), -signals do not interrupt NFS file operations. If -.B intr -is specified, system calls return EINTR if an in-progress NFS operation is interrupted by -a signal. -.IP -Using the -.B intr -option is preferred to using the -.B soft -option because it is significantly less likely to result in data corruption. -.IP -The -.BR intr " / " nointr -mount option is deprecated after kernel 2.6.25. -Only SIGKILL can interrupt a pending NFS operation on these kernels, -and if specified, this mount option is ignored to provide backwards -compatibility with older kernels. -.TP 1.5i .BR cto " / " nocto Selects whether to use close-to-open cache coherence semantics. If neither option is specified (or if @@ -807,29 +789,6 @@ The mount request fails if the server's rpcbind service is not available, the server's NFS service is not registered with its rpcbind service, or the server's NFS service is not available on the advertised port. .TP 1.5i -.BR intr " / " nointr -Selects whether to allow signals to interrupt file operations -on this mount point. If neither option is specified (or if -.B intr -is specified), system calls return EINTR if an in-progress NFS operation -is interrupted by a signal. If -.B nointr -is specified, signals do not -interrupt NFS operations. -.IP -Using the -.B intr -option is preferred to using the -.B soft -option because it is significantly less likely to result in data corruption. -.IP -The -.BR intr " / " nointr -mount option is deprecated after kernel 2.6.25. -Only SIGKILL can interrupt a pending NFS operation on these kernels, -and if specified, this mount option is ignored to provide backwards -compatibility with older kernels. -.TP 1.5i .BR cto " / " nocto Selects whether to use close-to-open cache coherence semantics for NFS directories on this mount point.
Signed-off-by: Jim Rees <rees@umich.edu> --- utils/mount/nfs.man | 53 ++++++----------------------------------------------- 1 file changed, 6 insertions(+), 47 deletions(-)