[LinuxBIOS] New hdama commit

[Jason Schildt]

All,

I've just made a commit up to the public repository from the Linux
Networx repository mirror.  The changes only effect Opteron single/dual
core and hdama board specific code.
Regards,

--jason--

[Jonathan McDowell]

On Tue, Aug 09, 2005 at 04:04:38PM -0600, Jason Schildt wrote:
> I've just made a commit up to the public repository from the Linux
> Networx repository mirror.  The changes only effect Opteron single/dual
> core and hdama board specific code.

Untrue. Your commit also touched romcc.c and appears to have broken
every build except tyan-s2735.

Compare:

http://the.earth.li/~noodles/lb/2003/log.html
with
http://the.earth.li/~noodles/lb/2004/log.html

J.

[yhlu]

1. write_irq_tables changes is overwrite by you. that will broke all
other Opteron MB. Please change that back.

YH

On 8/9/05, Jason Schildt  wrote:
> All,
>
> I've just made a commit up to the public repository from the Linux
> Networx repository mirror.  The changes only effect Opteron single/dual
> core and hdama board specific code.
> Regards,
>
> --jason--
>
> Jason W. Schildt
> LinuxBIOS Software Engineer
> Linux Networx
>
> LinuxBIOS mailing list
> [e-mail]
> http://www.openbios.org/mailman/listinfo/linuxbios

[yhlu]

Modified: trunk/LinuxBIOSv2/src/arch/i386/boot/tables.c
===================================================================
--- trunk/LinuxBIOSv2/src/arch/i386/boot/tables.c       2005-08-08
08:16:23 UTC (rev 2003)
+++ trunk/LinuxBIOSv2/src/arch/i386/boot/tables.c       2005-08-09
21:53:07 UTC (rev 2004)
@@ -23,7 +23,7 @@
       post_code(0x9a);

/* This table must be betweeen 0xf0000 & 0x100000 */
-       rom_table_end = write_pirq_routing_table(rom_table_end);
+       rom_table_end = copy_pirq_routing_table(rom_table_end);
       rom_table_end = (rom_table_end + 1023) & ~1023;

/* copy the smp block to address 0 */

On 8/9/05, yhlu  wrote:
> write_irq_tables changes is overwrite by you. that will broke all
>
> other Opteron MB. Please change that back.
>
> YH
>
> On 8/9/05, Jason Schildt  wrote:
> > All,
> >
> > I've just made a commit up to the public repository from the Linux
> > Networx repository mirror.  The changes only effect Opteron single/dual
> > core and hdama board specific code.
> > Regards,
> >
> > --jason--
> >
> > Jason W. Schildt
> > LinuxBIOS Software Engineer
> > Linux Networx
> >
> > LinuxBIOS mailing list
> > [e-mail]
> > http://www.openbios.org/mailman/listinfo/linuxbios

[yhlu]

2. you overwite the enable ext id and apic id lifting

YH

On 8/9/05, yhlu  wrote:
> write_irq_tables changes is overwrite by you. that will broke all
>
> other Opteron MB. Please change that back.
>
> YH
>
> On 8/9/05, Jason Schildt  wrote:
> > All,
> >
> > I've just made a commit up to the public repository from the Linux
> > Networx repository mirror.  The changes only effect Opteron single/dual
> > core and hdama board specific code.
> > Regards,
> >
> > --jason--
> >
> > Jason W. Schildt
> > LinuxBIOS Software Engineer
> > Linux Networx
> >
> > LinuxBIOS mailing list
> > [e-mail]
> > http://www.openbios.org/mailman/listinfo/linuxbios

[yhlu]

3. are you sure the node1/core0 could start node1/core1?
Accorinding to my testing last year, node1/core1 only can be started
by node0/core0.

YH

On 8/9/05, yhlu  wrote:
> you overwite the enable ext id and apic id lifting
>
> YH
>
> On 8/9/05, yhlu  wrote:
> > write_irq_tables changes is overwrite by you. that will broke all
> >
> > other Opteron MB. Please change that back.
> >
> > YH
> >
> > On 8/9/05, Jason Schildt  wrote:
> > > All,
> > >
> > > I've just made a commit up to the public repository from the Linux
> > > Networx repository mirror.  The changes only effect Opteron single/dual
> > > core and hdama board specific code.
> > > Regards,
> > >
> > > --jason--
> > >
> > > Jason W. Schildt
> > > LinuxBIOS Software Engineer
> > > Linux Networx
> > >
> > > LinuxBIOS mailing list
> > > [e-mail]
> > > http://www.openbios.org/mailman/listinfo/linuxbios

[yhlu]

On 8/9/05, yhlu  wrote:
> are you sure the node1/core0 could start node1/core1?
>
> Accorinding to my testing last year, node1/core1 only can be started
> by node0/core0.
>
> YH

So you move "adding core1 to cpu_bus" from cpu_bus_scan to amd_sibling_init.

the result will be when you access the cpu_bus and call
amd_sibling_init, the amd_sibling_init will alter the cpu_bus and
append ...cpu on that. So if for 4 way dual core or 8 way dual core,
when disable CPU_INIT_SERIALIZE, if it could cause racing...

YH

[yhlu]

4. it's good, to move hardware mem hole support from raminit.c to
northbridge.c, So it will be dynamically. but problem is it will use
more MTRRs. So carry_on should be calculated in some way...

YH

On 8/9/05, yhlu  wrote:
> are you sure the node1/core0 could start node1/core1?
>
> Accorinding to my testing last year, node1/core1 only can be started
> by node0/core0.
>
> YH
>
> On 8/9/05, yhlu  wrote:
> > you overwite the enable ext id and apic id lifting
> >
> > YH
> >
> > On 8/9/05, yhlu  wrote:
> > > write_irq_tables changes is overwrite by you. that will broke all
> > >
> > > other Opteron MB. Please change that back.
> > >
> > > YH
> > >
> > > On 8/9/05, Jason Schildt  wrote:
> > > > All,
> > > >
> > > > I've just made a commit up to the public repository from the Linux
> > > > Networx repository mirror.  The changes only effect Opteron single/dual
> > > > core and hdama board specific code.
> > > > Regards,
> > > >
> > > > --jason--
> > > >
> > > > Jason W. Schildt
> > > > LinuxBIOS Software Engineer
> > > > Linux Networx
> > > >
> > > > LinuxBIOS mailing list
> > > > [e-mail]
> > > > http://www.openbios.org/mailman/listinfo/linuxbios

[yhlu]

5. 
+                       /* Enable extended apic id's and second core */
+                       val = pci_read_config32(dev_f0, 0x68);
+                       val |= (1 << 18) | (1 << 17) | ( 1 << 5);
+                       pci_write_config32(dev_f0, 0x68, val);

ext apic id mode is only need to enable for 8 way dual core system.

6.  incohere_ht.c overwrite my new changes need to be roll back.
7. debug.c need to rollback.

YH

On 8/9/05, yhlu  wrote:
> it's good, to move hardware mem hole support from raminit.c to
>
> northbridge.c, So it will be dynamically. but problem is it will use
> more MTRRs. So carry_on should be calculated in some way...
>
> YH
>
> On 8/9/05, yhlu  wrote:
> > are you sure the node1/core0 could start node1/core1?
> >
> > Accorinding to my testing last year, node1/core1 only can be started
> > by node0/core0.
> >
> > YH
> >
> > On 8/9/05, yhlu  wrote:
> > > you overwite the enable ext id and apic id lifting
> > >
> > > YH
> > >
> > > On 8/9/05, yhlu  wrote:
> > > > write_irq_tables changes is overwrite by you. that will broke all
> > > >
> > > > other Opteron MB. Please change that back.
> > > >
> > > > YH
> > > >
> > > > On 8/9/05, Jason Schildt  wrote:
> > > > > All,
> > > > >
> > > > > I've just made a commit up to the public repository from the Linux
> > > > > Networx repository mirror.  The changes only effect Opteron single/dual
> > > > > core and hdama board specific code.
> > > > > Regards,
> > > > >
> > > > > --jason--
> > > > >
> > > > > Jason W. Schildt
> > > > > LinuxBIOS Software Engineer
> > > > > Linux Networx
> > > > >
> > > > > LinuxBIOS mailing list
> > > > > [e-mail]
> > > > > http://www.openbios.org/mailman/listinfo/linuxbios

[BULK] Re: [LinuxBIOS] New hdama commit

[Jason Schildt]

Oh dear. I've changed it back to the original.  Our build works with
both, so no loss on our side.
My apologies.

--jason--

On Tue, Aug 09, 2005 at 11:53:51PM +0100, Jonathan McDowell wrote:
> On Tue, Aug 09, 2005 at 04:04:38PM -0600, Jason Schildt wrote:
> > I've just made a commit up to the public repository from the Linux
> > Networx repository mirror.  The changes only effect Opteron single/dual
> > core and hdama board specific code.
>
> Untrue. Your commit also touched romcc.c and appears to have broken
> every build except tyan-s2735.
>
> Compare:
>
> http://the.earth.li/~noodles/lb/2003/log.html
> with
> http://the.earth.li/~noodles/lb/2004/log.html
>
> J.
>
> LinuxBIOS mailing list
> [e-mail]
> http://www.openbios.org/mailman/listinfo/linuxbios

[yhlu]

the init_ecc part need to skip the mem_hole region, You need put back
several line in init_ecc.

YH

On 8/9/05, yhlu  wrote:
> it's good, to move hardware mem hole support from raminit.c to
>
> northbridge.c, So it will be dynamically. but problem is it will use
> more MTRRs. So carry_on should be calculated in some way...
>
> YH
>
> On 8/9/05, yhlu  wrote:
> > are you sure the node1/core0 could start node1/core1?
> >
> > Accorinding to my testing last year, node1/core1 only can be started
> > by node0/core0.
> >
> > YH
> >
> > On 8/9/05, yhlu  wrote:
> > > you overwite the enable ext id and apic id lifting
> > >
> > > YH
> > >
> > > On 8/9/05, yhlu  wrote:
> > > > write_irq_tables changes is overwrite by you. that will broke all
> > > >
> > > > other Opteron MB. Please change that back.
> > > >
> > > > YH
> > > >
> > > > On 8/9/05, Jason Schildt  wrote:
> > > > > All,
> > > > >
> > > > > I've just made a commit up to the public repository from the Linux
> > > > > Networx repository mirror.  The changes only effect Opteron single/dual
> > > > > core and hdama board specific code.
> > > > > Regards,
> > > > >
> > > > > --jason--
> > > > >
> > > > > Jason W. Schildt
> > > > > LinuxBIOS Software Engineer
> > > > > Linux Networx
> > > > >
> > > > > LinuxBIOS mailing list
> > > > > [e-mail]
> > > > > http://www.openbios.org/mailman/listinfo/linuxbios