Re: Linux 2.4.34-pre3

Re: Linux 2.4.34-pre3

[Mikael Pettersson]

On Tue, 19 Sep 2006 17:32:53 +0000, Willy Tarreau wrote:
>I've been a little bit silent and quite busy too. As announced with -pre2,
>here comes -pre3 with only GCC4 fixes. Other fixes I received are minor
>and can wait for -pre4. I really want people to test -pre3 without adding
>any noise to the test. There should be *no* regression at all with existing
>compilers.

I've reviewed the changes and they look fine. So far 2.4.34-pre3
compiled with gcc-4.1.1 runs on i386, x86_64, and ppc32 without
any regressions. I'll test 2.4.34-pre3 on sparc64 this weekend,
but since 2.4.34-pre2 + my gcc4 fixes already runs fine on sparc64,
I don't expect any problems.

/Mikael

Re: Linux 2.4.34-pre3

[Willy Tarreau]

On Wed, Sep 20, 2006 at 04:51:41PM +0200, Mikael Pettersson wrote:
> On Tue, 19 Sep 2006 17:32:53 +0000, Willy Tarreau wrote:
> >I've been a little bit silent and quite busy too. As announced with -pre2,
> >here comes -pre3 with only GCC4 fixes. Other fixes I received are minor
> >and can wait for -pre4. I really want people to test -pre3 without adding
> >any noise to the test. There should be *no* regression at all with existing
> >compilers.
> 
> I've reviewed the changes and they look fine. So far 2.4.34-pre3
> compiled with gcc-4.1.1 runs on i386, x86_64, and ppc32 without
> any regressions. I'll test 2.4.34-pre3 on sparc64 this weekend,
> but since 2.4.34-pre2 + my gcc4 fixes already runs fine on sparc64,
> I don't expect any problems.

Thanks for your tests Mikael. FYI, it works fine for me on my ultra5
(sparc64 UP) and ultra60 (sparc64 SMP).

> /Mikael

Regards,
Willy

Linux 2.4.34-pre3

[Willy Tarreau]

Hello !

I've been a little bit silent and quite busy too. As announced with -pre2,
here comes -pre3 with only GCC4 fixes. Other fixes I received are minor
and can wait for -pre4. I really want people to test -pre3 without adding
any noise to the test. There should be *no* regression at all with existing
compilers.

Gcc 4.1 is known to build x86, x86_64, ppc, sparc64, and sparc. Only sparc
has received no testing yet, while sparc64 is OK. It is possible that some
sparc/sparc64 drivers have not been caught because of undetected options
combinations (Davem CC'd for any possible advice on this mater). Status
for other archs is unknown but at least must not be affected for existing
setups.

You will notice that most of the changes below appear under my name while
I do not deserve any credit for the changes. It is just because I've cut
all the fixes to sort them, and committed them myself individually. But
the real work has been done by Mikael.

We have worked *very* carefully on this merge, and we hope to get all
possible feedback. People who encounter build problems on archs listed
above are free to report them, possibly with the fix. When providing a
fix, *please* provide the whole error output in the commit message so
that we can track what has been fixed. People who want to include support
for other archs will have to provide patches, as (at least for me) we
are not equipped to build on other archs (except for alpha when my RAM
arrives).

I plan to wait up to the end of this month before providing -pre4 if there
is no feedback. Important fixes will be subject to another -stable release
anyway, so it's safe to wait for feedback here.

Now I've fixed my release scripts, so the changelog and patch should be
OK ;-)

Best regards,
Willy


Summary of changes from v2.4.34-pre2 to v2.4.34-pre3
============================================

Mikael Pettersson:
      [GCC4] SPARC64: fix UP build error in arch/sparc64/mm/init.c

Willy Tarreau:
      [GCC4] add preliminary support for GCC 4 (Mikael Pettersson)
      [GCC4] fix build error in include/linux/generic_serial.h
      [GCC4] fix build error in include/net/irda/qos.h
      [GCC4] fix build error in include/linux/fsfilter.h
      [GCC4] fix build error in include/linux/intermezzo_fs.h
      [GCC4] fix build error in include/net/udp.h
      [GCC4] fix build error in include/net/irda/irttp.h
      [GCC4] fix build error in include/net/irda/irlan_event.h
      [GCC4] fix build error in include/asm-ppc/spinlock.h
      [GCC4] fix build error in fs/intermezzo/presto.c
      [GCC4] fix build error in net/ipv6/ip6_fib.c
      [GCC4] fix build error in net/ipv6/sysctl_net_ipv6.c
      [GCC4] fix build error in net/khttpd/prototypes.h
      [GCC4] fix build error in drivers/block/nbd.c
      [GCC4] fix build error in drivers/block/xd.c
      [GCC4] fix build error in drivers/block/paride/pd.c
      [GCC4] fix build error in drivers/char/sonypi.h
      [GCC4] fix build error in drivers/char/sonypi.h
      [GCC4] fix build error in drivers/char/tpqic02.c
      [GCC4] fix build error in drivers/char/drm-4.0/drmP.h
      [GCC4] fix build error in drivers/char/rio/rio_linux.c
      [GCC4] fix build error in drivers/net/acenic.c
      [GCC4] fix build error in drivers/net/wan/comx.h
      [GCC4] fix build error in drivers/net/3c507.c
      [GCC4] fix build error in drivers/net/arlan.c
      [GCC4] fix build error in drivers/net/irda/donauboe.c
      [GCC4] fix build error in drivers/net/sk98lin/skvpd.c
      [GCC4] fix build error in drivers/net/wan/comx-hw-comx.c
      [GCC4] fix build error in drivers/net/wan/sdladrv.c
      [GCC4] fix build error in drivers/net/wan/sdlamain.c
      [GCC4] fix build error in drivers/net/wan/sdla_fr.c
      [GCC4] fix build error in drivers/net/hamradio/baycom_epp.c
      [GCC4] fix build error in drivers/net/hamradio/soundmodem/sm.h
      [GCC4] fix build error in drivers/scsi/advansys.c
      [GCC4] fix build error in drivers/scsi/atp870u.c
      [GCC4] fix build error in drivers/scsi/cpqfcTS*
      [GCC4] fix build error in drivers/ide/legacy/hd.c
      [GCC4] fix build error in drivers/cdrom/sbpcd.c
      [GCC4] fix build error in drivers/md/lvm-internal.h
      [GCC4] fix build error in drivers/atm/iphase.c
      [GCC4] fix build error in drivers/atm/fore200e.c
      [GCC4] fix build error in drivers/isdn/eicon/eicon.h
      [GCC4] fix build error in drivers/isdn/hisax/hfc_pci.c
      [GCC4] fix build error in drivers/i2c/i2c-core.c
      [GCC4] fix build error in drivers/i2c/i2c-proc.c
      [GCC4] fix build error in drivers/media/video/videodev.c
      [GCC4] fix build error in drivers/usb/audio.c
      [GCC4] fix build error in drivers/ieee1394/highlevel.c
      [GCC4] fix build error in drivers/media/video/bttvp.h
      [GCC4] fix build error in drivers/sound/wavfront.c
      [GCC4] fix warning in include/linux/atalk.h
      [GCC4] fix warnings in include/linux/isdnif.h
      [GCC4] fix warnings in include/net/dn_dev.h
      [GCC4] fix warnings in include/net/dn_nsp.h
      [GCC4] fix warnings in sdla.h and if_frad.h
      [GCC4] fix warnings in sdla_x25.c and sdla_x25.h
      [GCC4] fix warnings in include/linux/wanpipe.h
      [GCC4] fix warnings in drivers/char/sx.c
      [GCC4] fix warning in drivers/char/ip2/i2lib.c
      [GCC4] fix warnings in drivers/net/de4x5,depca,arcnet
      [GCC4] fix warnings in drivers/isdn/eicon/eicon*.h
      [GCC4] fix warnings in drivers/isdn/hisax/hisax.h
      [GCC4] fix build in drivers/atm/horizon.c
      [GCC4] fix build error in drivers/net/rrunner.c
      [GCC4] SPARC64: fix build error in arch/sparc64/kernel/smp.c
      [GCC4] SPARC64: fix build error in arch/sparc64/kernel/time.c
      [GCC4] SPARC64: fix build error in drivers/sbus/char/pcikbd.c
      [GCC4] SPARC: fix build error in arch/sparc/kernel/signal.c
      [GCC4] SPARC: fix build error in arch/sparc/kernel/time.c
      [GCC4] SPARC: fix build error in drivers/fc4/soc.c
      Merge branch 'gcc4'
      Change VERSION to 2.4.34-pre3

Re: Linux 2.4.34-pre3

[Nick Piggin]

Willy Tarreau wrote:
> Hello !
> 
> I've been a little bit silent and quite busy too. As announced with -pre2,
> here comes -pre3 with only GCC4 fixes. Other fixes I received are minor
> and can wait for -pre4. I really want people to test -pre3 without adding
> any noise to the test. There should be *no* regression at all with existing
> compilers.
> 
> Gcc 4.1 is known to build x86, x86_64, ppc, sparc64, and sparc. Only sparc
> has received no testing yet, while sparc64 is OK. It is possible that some
> sparc/sparc64 drivers have not been caught because of undetected options
> combinations (Davem CC'd for any possible advice on this mater). Status
> for other archs is unknown but at least must not be affected for existing
> setups.
> 
> You will notice that most of the changes below appear under my name while
> I do not deserve any credit for the changes. It is just because I've cut
> all the fixes to sort them, and committed them myself individually. But
> the real work has been done by Mikael.
> 
> We have worked *very* carefully on this merge, and we hope to get all
> possible feedback. People who encounter build problems on archs listed
> above are free to report them, possibly with the fix. When providing a
> fix, *please* provide the whole error output in the commit message so
> that we can track what has been fixed. People who want to include support
> for other archs will have to provide patches, as (at least for me) we
> are not equipped to build on other archs (except for alpha when my RAM
> arrives).
> 
> I plan to wait up to the end of this month before providing -pre4 if there
> is no feedback. Important fixes will be subject to another -stable release
> anyway, so it's safe to wait for feedback here.

I wonder if 2.4 doesn't need the memory ordering fix to prevent pagecache
corruption in reclaim? (http://www.gatago.com/linux/kernel/14682626.html)

What would need to be done is to test page_count before testing PageDirty,
and putting an smp_rmb between the two.

-- 
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 

Re: Linux 2.4.34-pre3

[Willy Tarreau]

Hi Nick,

On Wed, Sep 20, 2006 at 03:42:06AM +1000, Nick Piggin wrote:

[cut -pre3 advertisement]

> I wonder if 2.4 doesn't need the memory ordering fix to prevent pagecache
> corruption in reclaim? (http://www.gatago.com/linux/kernel/14682626.html)
> 
> What would need to be done is to test page_count before testing PageDirty,
> and putting an smp_rmb between the two.

I've read the thread, and Linus proposed to add an smp_wmb() in
set_page_dirty() too. I see that an smp_rmb() is already present
in shrink_cache() with the adequate comment. set_page_dirty() begins
with a test_and_set_bit() check. I suspect that transposing the fix
for 2.6 to 2.4 would imply to and an smp_wmb() here :

void fastcall set_page_dirty(struct page *page)
{
        if (!test_and_set_bit(PG_dirty, &page->flags)) {
                struct address_space *mapping = page->mapping;

+		smp_wmb();
                if (mapping) {
                        spin_lock(&pagecache_lock);
                        mapping = page->mapping;

But I'm really reluctant on this change, as my knowledge here is rather
limited. Also, the fact that the first proposed part of the patch is here
makes me think that it has already been considered OK.

Maybe I'm wrong, I've CC'd Marcelo for any comments.

Cheers,
Willy

Re: Linux 2.4.34-pre3

[Nick Piggin]

Willy Tarreau wrote:
> Hi Nick,
> 
> On Wed, Sep 20, 2006 at 03:42:06AM +1000, Nick Piggin wrote:
> 
> [cut -pre3 advertisement]
> 
> 
>>I wonder if 2.4 doesn't need the memory ordering fix to prevent pagecache
>>corruption in reclaim? (http://www.gatago.com/linux/kernel/14682626.html)
>>
>>What would need to be done is to test page_count before testing PageDirty,
>>and putting an smp_rmb between the two.
> 
> 
> I've read the thread, and Linus proposed to add an smp_wmb() in
> set_page_dirty() too.

I think that isn't needed because put_page is a RMW, which is defined
to order memory. And presumably you wouldn't set the page dirty without
a reference to the page.

> I see that an smp_rmb() is already present
> in shrink_cache() with the adequate comment.

So there is! My mistake then, I was confused and looking at
try_to_swap_out, but I see that doesn't actually free the page. Fine,
I think 2.4 is OK then.

-- 
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 

Re: Linux 2.4.34-pre3

[Willy Tarreau]

On Wed, Sep 20, 2006 at 04:55:25AM +1000, Nick Piggin wrote:
> Willy Tarreau wrote:
> >Hi Nick,
> >
> >On Wed, Sep 20, 2006 at 03:42:06AM +1000, Nick Piggin wrote:
> >
> >[cut -pre3 advertisement]
> >
> >
> >>I wonder if 2.4 doesn't need the memory ordering fix to prevent pagecache
> >>corruption in reclaim? (http://www.gatago.com/linux/kernel/14682626.html)
> >>
> >>What would need to be done is to test page_count before testing PageDirty,
> >>and putting an smp_rmb between the two.
> >
> >
> >I've read the thread, and Linus proposed to add an smp_wmb() in
> >set_page_dirty() too.
> 
> I think that isn't needed because put_page is a RMW, which is defined
> to order memory. And presumably you wouldn't set the page dirty without
> a reference to the page.

OK, thanks for the explanation.

> >I see that an smp_rmb() is already present
> >in shrink_cache() with the adequate comment.
> 
> So there is! My mistake then, I was confused and looking at
> try_to_swap_out, but I see that doesn't actually free the page. Fine,
> I think 2.4 is OK then.

Perfect !

Thanks,
Willy