curious about BK checkin protocol

curious about BK checkin protocol

[Robert P. J. Day]

  i just bk-cloned a fresh copy of the source tree from
http://ppc.bkbits.net:8080/linuxppc-2.5, and once again, had to fix the
file arch/ppc/kernel/ppc_ksyms.c to remove the now-obsolete snippet of
code:

#if defined(CONFIG_8xx)
EXPORT_SYMBOL(request_8xxirq);
#endif

  i thought it had been well-established by now that this had to go.
what's the protocol for someone putting these changes into the tree?
just curious.

rday

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: curious about BK checkin protocol

[Matt Porter]

On Tue, May 25, 2004 at 09:12:10AM -0400, Robert P. J. Day wrote:
>
>   i just bk-cloned a fresh copy of the source tree from
> http://ppc.bkbits.net:8080/linuxppc-2.5, and once again, had to fix the
> file arch/ppc/kernel/ppc_ksyms.c to remove the now-obsolete snippet of
> code:
>
> #if defined(CONFIG_8xx)
> EXPORT_SYMBOL(request_8xxirq);
> #endif
>
>   i thought it had been well-established by now that this had to go.
> what's the protocol for someone putting these changes into the tree?
> just curious.

Post a patch.  If it's something that is incorrect in linux-2.5
as well, then the patch is expected to be against linux-2.5.

-Matt

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

405ep ether port broken in performance test.

[Davey]

I am developing a firewall based on ibm405ep, it have 4 ether ports, two on-chip and two 8139c. And it run 2.4.18 MontaVista kernel.
Between two on-chip ports, one of a port will be broken after 0-2 seconds when I test the transaction rate with small packet (64 byte). and the port will return to fine 2-5minutes after the test end. When I use big packets, i works very fine. My tool is netiq chariot.

Between on-chip port and 8139c port, the result is fine.

Can anyone  give me some help?

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: curious about BK checkin protocol

[Robert P. J. Day]

On Tue, 25 May 2004, Matt Porter wrote:

> On Tue, May 25, 2004 at 09:12:10AM -0400, Robert P. J. Day wrote:
> >
> >   i just bk-cloned a fresh copy of the source tree from
> > http://ppc.bkbits.net:8080/linuxppc-2.5, and once again, had to fix the
> > file arch/ppc/kernel/ppc_ksyms.c to remove the now-obsolete snippet of
> > code:
> >
> > #if defined(CONFIG_8xx)
> > EXPORT_SYMBOL(request_8xxirq);
> > #endif
> >
> >   i thought it had been well-established by now that this had to go.
> > what's the protocol for someone putting these changes into the tree?
> > just curious.
>
> Post a patch.  If it's something that is incorrect in linux-2.5
> as well, then the patch is expected to be against linux-2.5.

post to this list?  sure, if that's the right place.  the only reason i'm
obsessed about that little fix as opposed to all the others that are going
in is that, WRT the most recent BK pull from
http://ppc.bkbits.net:8080/linuxppc-2.5, that's the *only* thing that
keeps the kernel from compiling, and letting me build that a kernel that,
while it loads and runs, admittedly still blows up upon starting init.

while i realize that the current kernel still has user land problems, it
seems a shame to not at least fix the single minor thing that prevents a
simple build.

rday


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: curious about BK checkin protocol

[Tom Rini]

On Wed, May 26, 2004 at 07:19:53AM -0400, Robert P. J. Day wrote:

> On Tue, 25 May 2004, Matt Porter wrote:
> > On Tue, May 25, 2004 at 09:12:10AM -0400, Robert P. J. Day wrote:
> > >   i just bk-cloned a fresh copy of the source tree from
> > > http://ppc.bkbits.net:8080/linuxppc-2.5, and once again, had to fix the
> > > file arch/ppc/kernel/ppc_ksyms.c to remove the now-obsolete snippet of
> > > code:
> > >
> > > #if defined(CONFIG_8xx)
> > > EXPORT_SYMBOL(request_8xxirq);
> > > #endif
> > >
> > >   i thought it had been well-established by now that this had to go.
> > > what's the protocol for someone putting these changes into the tree?
> > > just curious.
> >
> > Post a patch.  If it's something that is incorrect in linux-2.5
> > as well, then the patch is expected to be against linux-2.5.
>
> post to this list?  sure, if that's the right place.  the only reason i'm
> obsessed about that little fix as opposed to all the others that are going
> in is that, WRT the most recent BK pull from
> http://ppc.bkbits.net:8080/linuxppc-2.5, that's the *only* thing that
> keeps the kernel from compiling, and letting me build that a kernel that,
> while it loads and runs, admittedly still blows up upon starting init.
>
> while i realize that the current kernel still has user land problems, it
> seems a shame to not at least fix the single minor thing that prevents a
> simple build.

I still like to argue that it's best to loudly blow up than to compile
fine and then die in other ways at run-time (if someone hadn't gotten
the last problem fixed on 8xx I was getting tempted to move what we had
done now up with an #error tossed on top of head_8xx.S pointing people
to what's wrong, etc).

--
Tom Rini
http://gate.crashing.org/~trini/

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: curious about BK checkin protocol

[Robert P. J. Day]

On Wed, 26 May 2004, Tom Rini wrote:

> I still like to argue that it's best to loudly blow up than to compile
> fine and then die in other ways at run-time (if someone hadn't gotten
> the last problem fixed on 8xx I was getting tempted to move what we had
> done now up with an #error tossed on top of head_8xx.S pointing people
> to what's wrong, etc).

that's a fair statement.  but, from my perspective, before i tried this
recently, i had *no* clue about the possibility of linux on 8xx, and
grabbed the linuxppc-2.5 tree, which of course didn't compile because of
that one issue.  once i got that compile issue out of the way, i was
purely deeeelighted to see a kernel actually load and boot to user space
and at least start init before hanging.  and this was after doing
*nothing* in terms of customizing the kernel for my board.

even with going no further than that, i could at least take comfort in the
fact that *something* was happening.  it's an admittedly nitpicky
perspective, but seeing even that partial progress made me one happy
camper.

anyway, at this point, i'll just be waiting with bated breath for new
patches.  too exciting for words ...

rday

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: curious about BK checkin protocol

[Matt Porter]

On Wed, May 26, 2004 at 07:19:53AM -0400, Robert P. J. Day wrote:
>
>
> On Tue, 25 May 2004, Matt Porter wrote:
>
> > On Tue, May 25, 2004 at 09:12:10AM -0400, Robert P. J. Day wrote:
> > >   i thought it had been well-established by now that this had to go.
> > > what's the protocol for someone putting these changes into the tree?
> > > just curious.
> >
> > Post a patch.  If it's something that is incorrect in linux-2.5
> > as well, then the patch is expected to be against linux-2.5.
>
> post to this list?  sure, if that's the right place.  the only reason i'm
> obsessed about that little fix as opposed to all the others that are going

Yes, post to this list for embedded-specific patches (8xx). If
it's a clear fix, somebody will put it into linux-2.5 and it
will be pulled into linuxppc-2.5 on the next merge. Put [PATCH]
in the subject and mention what BK tree it's made against.

Let me say this again, but a bit clearer so that _everybody_
understands.  A patch is the standard unit of communication here.
If you don't post a patch that represents your ideas or proposed
fixes, your thoughts are likely to get lost in all the noise on
these lists.

-Matt

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/