Re: 2.6.25-rc7-git2: Reported regressions from 2.6.24

Re: 2.6.25-rc7-git2: Reported regressions from 2.6.24

[Andrew Morton]

On Fri, 28 Mar 2008 11:48:55 +0100 Pawe__ Staszewski <pstaszewski@artcom.pl> wrote:

> Linus Torvalds pisze:
> > On Thu, 27 Mar 2008, Christoph Lameter wrote:
> >   
> >> Slab allocations can never use GFP_HIGHMEM.
> >>     
> >
> > Totally irrelevant.
> >
> > The page allocation path does
> >
> >         if (gfp_flags & __GFP_ZERO)
> >                 prep_zero_page(page, order, gfp_flags);
> >
> > and that will cause a warning REGARDLESS of whether the page is a HIGHMEM 
> > page or not.
> >
> > And the fact is, passing in GFP_ZERO from the SLUB code is a bug 
> > regardless, because it unnecessarily does the dual memset().
> >
> > So here's a damn big clue:
> >  - SLUB does its own GFP_ZERO handling
> >  - so passing GFP_ZERO down to the page allocator is a f*cking bug
> >  - and this has NOTHING what-so-ever to do with GFP_HIGHMEM or even 
> >    whether the warning is "valid" or not - it's a bug even if the warning 
> >    had never happened.
> >
> > So stop blathering, and just admit that this was buggy. It was also 
> > fundamentally fragile to leave GFP_ZERO around when it was known to not be 
> > valid at that point (exactly because GFP_ZERO was handled by the caller).
> >
> > 		Linus
> >
> >
> >   
> Sorry for offtopic but i have the same problem with kernels 2.6.25-*
> like:
> http://kerneltrap.org/mailarchive/linux-netdev/2008/3/27/1274804
> http://kerneltrap.org/mailarchive/linux-netdev/2008/3/27/1270334
> 
> I search linux-netdev and found this links.
> I only sugest that 
> 
> 	Denys Fedoryshchenko
> 
> can have the same problem that i have with this kernels.
> I must revert my all kernels to 2.6.23.11 to get stable work on high (ip 
> traffic) loads.
> 
> And there is no documentation for LRO... and Stephen Hemminger write me 
> that LRO is not compatible with bridgeing and routing.
> see this link:
> http://bugzilla.kernel.org/show_bug.cgi?id=10335
> 
> 
> So there must be some documentation for this ... because people can have 
> many problems with this.
> 

These are all networking things, so let's cc that list.

> 
> Sorry for offtopic but this can resolve problems like my and Denys .

It's very on-topic - thanks for the reminder.

Rafael, are these things actually on the list?

Re: 2.6.25-rc7-git2: Reported regressions from 2.6.24

[Rafael J. Wysocki]

On Friday, 28 of March 2008, Andrew Morton wrote:
> On Fri, 28 Mar 2008 11:48:55 +0100 Pawe__ Staszewski <pstaszewski@artcom.pl> wrote:
> 
> > Linus Torvalds pisze:
> > > On Thu, 27 Mar 2008, Christoph Lameter wrote:
> > >   
> > >> Slab allocations can never use GFP_HIGHMEM.
> > >>     
> > >
> > > Totally irrelevant.
> > >
> > > The page allocation path does
> > >
> > >         if (gfp_flags & __GFP_ZERO)
> > >                 prep_zero_page(page, order, gfp_flags);
> > >
> > > and that will cause a warning REGARDLESS of whether the page is a HIGHMEM 
> > > page or not.
> > >
> > > And the fact is, passing in GFP_ZERO from the SLUB code is a bug 
> > > regardless, because it unnecessarily does the dual memset().
> > >
> > > So here's a damn big clue:
> > >  - SLUB does its own GFP_ZERO handling
> > >  - so passing GFP_ZERO down to the page allocator is a f*cking bug
> > >  - and this has NOTHING what-so-ever to do with GFP_HIGHMEM or even 
> > >    whether the warning is "valid" or not - it's a bug even if the warning 
> > >    had never happened.
> > >
> > > So stop blathering, and just admit that this was buggy. It was also 
> > > fundamentally fragile to leave GFP_ZERO around when it was known to not be 
> > > valid at that point (exactly because GFP_ZERO was handled by the caller).
> > >
> > > 		Linus
> > >
> > >
> > >   
> > Sorry for offtopic but i have the same problem with kernels 2.6.25-*
> > like:
> > http://kerneltrap.org/mailarchive/linux-netdev/2008/3/27/1274804
> > http://kerneltrap.org/mailarchive/linux-netdev/2008/3/27/1270334
> > 
> > I search linux-netdev and found this links.
> > I only sugest that 
> > 
> > 	Denys Fedoryshchenko
> > 
> > can have the same problem that i have with this kernels.
> > I must revert my all kernels to 2.6.23.11 to get stable work on high (ip 
> > traffic) loads.
> > 
> > And there is no documentation for LRO... and Stephen Hemminger write me 
> > that LRO is not compatible with bridgeing and routing.
> > see this link:
> > http://bugzilla.kernel.org/show_bug.cgi?id=10335
> > 
> > 
> > So there must be some documentation for this ... because people can have 
> > many problems with this.
> > 
> 
> These are all networking things, so let's cc that list.
> 
> > 
> > Sorry for offtopic but this can resolve problems like my and Denys .
> 
> It's very on-topic - thanks for the reminder.
> 
> Rafael, are these things actually on the list?

Well, this isn't a recent regression, at least not from 2.6.24, so I don't
list it.