Re: Yh Lu's patches

[YhLu]

Eric,

As for cpufixup.c, actually I grab the fragment from raminit.c's
set_top_ram.
I don't want to make my own, because I can not access the most update code
because of 24 hours delay.

Fill_last:
Can someone raise the bug to LinuxKernel developer team?

There are Config.lb and config.lb in the southbridge/amd/amd8111.
We should get rid of config.lb.

Regards

Yinghai Lu

-----邮件原件-----
发件人: [e-mail] [mailto:[e-mail]] 
发送时间: 2003年8月4日 16:35
收件人: ron minnich
抄送: [e-mail]
主题: Re: Yh Lu's patches

ron minnich  writes:
> Can Eric and Stefan and others interested please take a quick look at
> this, it is the non-tyan mods from Yh Lu.
>
> I am going to commit the tyan mods in a few minutes.
>
> thanks
>
> ron
> diff -uNr ./freebios2/src/cpu/k8/cpufixup.c

../freebios2/src/cpu/k8/cpufixup.c
> --- ./freebios2/src/cpu/k8/cpufixup.c	2003-08-04 17:57:43.000000000 -0400
> +++ ../freebios2/src/cpu/k8/cpufixup.c	2003-08-04

19:41:18.000000000 -0400
> @@ -29,15 +29,28 @@
>    	mmio_basek = tomk;
>    }
> /* Setup TOP_MEM */
> msr.hi = mmio_basek >> 22;
> msr.lo = mmio_basek << 10;
> wrmsr(TOP_MEM, msr);
>
> +#if 1
>    /* Report the amount of memory. */
>    print_debug("cpufixup RAM: 0x");
>    print_debug_hex32(tomk);
>    print_debug(" KB\r\n");
>
> +#endif
>    /* Now set top of memory */
>    msr.lo = (tomk & 0x003fffff) << 10;
>    msr.hi = (tomk & 0xffc00000) >> 22;
>    wrmsr(TOP_MEM2, msr);
>    /* Leave a 64M hole between TOP_MEM and TOP_MEM2
>     * so I can see my rom chip and other I/O devices.
>     */
>    if (tomk >= 0x003f0000) {
>            tomk = 0x3f0000;
>    } //    tom_k = 0x3c0000;
>    msr.lo = (tomk & 0x003fffff) << 10;
>    msr.hi = (tomk & 0xffc00000) >> 22;
>    wrmsr(TOP_MEM, msr);
>
> /* Setup TOP_MEM2 */
>
> msr.hi = tomk >> 22;
>
> msr.lo = tomk << 10;
>
> wrmsr(TOP_MEM2, msr);
>
> /* zero the IORR's before we enable to prevent
> undefined side effects.

This bit is redundant. 
I described the change to Yinghai Lu and he made is own version of it.

And in fact it is actively wrong. mmio_basek does a much better job
of leaving an I/O hole there, then simply subtracting 64MB from 4GB.
> diff -uNr ./freebios2/src/include/console/console.h
> ../freebios2/src/include/console/console.h
>
> --- ./freebios2/src/include/console/console.h 2003-07-19

00:28:21.000000000
> -0400
>
> +++ ../freebios2/src/include/console/console.h 2003-08-02

01:40:57.000000000
> -0400
>
> @@ -92,34 +92,34 @@
>  #define print_debug_char(CH)   printk_debug  ("%c", (CH))
>  #define print_spew_char(CH)    printk_spew   ("%c", (CH))
>
> -#define print_emerg_hex8(HEX)   printk_emerg  ("%08x",  (HEX))
> -#define print_alert_hex8(HEX)   printk_alert  ("%08x",  (HEX))
> -#define print_crit_hex8(HEX)    printk_crit   ("%08x",  (HEX))
> -#define print_err_hex8(HEX)     printk_err    ("%08x",  (HEX))
> -#define print_warning_hex8(HEX) printk_warning("%08x",  (HEX))
> -#define print_notice_hex8(HEX)  printk_notice ("%08x",  (HEX))
> -#define print_info_hex8(HEX)    printk_info   ("%08x",  (HEX))
> -#define print_debug_hex8(HEX)   printk_debug  ("%08x",  (HEX))
> -#define print_spew_hex8(HEX)    printk_spew   ("%08x",  (HEX))
> +#define print_emerg_hex8(HEX)   printk_emerg  ("%02x",  (HEX))
> +#define print_alert_hex8(HEX)   printk_alert  ("%02x",  (HEX))
> +#define print_crit_hex8(HEX)    printk_crit   ("%02x",  (HEX))
> +#define print_err_hex8(HEX)     printk_err    ("%02x",  (HEX))
> +#define print_warning_hex8(HEX) printk_warning("%02x",  (HEX))
> +#define print_notice_hex8(HEX)  printk_notice ("%02x",  (HEX))
> +#define print_info_hex8(HEX)    printk_info   ("%02x",  (HEX))
> +#define print_debug_hex8(HEX)   printk_debug  ("%02x",  (HEX))
> +#define print_spew_hex8(HEX)    printk_spew   ("%02x",  (HEX))
>
> -#define print_emerg_hex16(HEX)   printk_emerg  ("%016x", (HEX))
> -#define print_alert_hex16(HEX)   printk_alert  ("%016x", (HEX))
> -#define print_crit_hex16(HEX)    printk_crit   ("%016x", (HEX))
> -#define print_err_hex16(HEX)     printk_err    ("%016x", (HEX))
> -#define print_warning_hex16(HEX) printk_warning("%016x", (HEX))
> -#define print_notice_hex16(HEX)  printk_notice ("%016x", (HEX))
> -#define print_info_hex16(HEX)    printk_info   ("%016x", (HEX))
> -#define print_debug_hex16(HEX)   printk_debug  ("%016x", (HEX))
> -#define print_spew_hex16(HEX)    printk_spew   ("%016x", (HEX))
> +#define print_emerg_hex16(HEX)   printk_emerg  ("%04x", (HEX))
> +#define print_alert_hex16(HEX)   printk_alert  ("%04x", (HEX))
> +#define print_crit_hex16(HEX)    printk_crit   ("%04x", (HEX))
> +#define print_err_hex16(HEX)     printk_err    ("%04x", (HEX))
> +#define print_warning_hex16(HEX) printk_warning("%04x", (HEX))
> +#define print_notice_hex16(HEX)  printk_notice ("%04x", (HEX))
> +#define print_info_hex16(HEX)    printk_info   ("%04x", (HEX))
> +#define print_debug_hex16(HEX)   printk_debug  ("%04x", (HEX))
> +#define print_spew_hex16(HEX)    printk_spew   ("%04x", (HEX))
>
> -#define print_emerg_hex32(HEX)   printk_emerg  ("%032x", (HEX))
> -#define print_alert_hex32(HEX)   printk_alert  ("%032x", (HEX))
> -#define print_crit_hex32(HEX)    printk_crit   ("%032x", (HEX))
> -#define print_err_hex32(HEX)     printk_err    ("%032x", (HEX))
> -#define print_warning_hex32(HEX) printk_warning("%032x", (HEX))
> -#define print_notice_hex32(HEX)  printk_notice ("%032x", (HEX))
> -#define print_info_hex32(HEX)    printk_info   ("%032x", (HEX))
> -#define print_debug_hex32(HEX)   printk_debug  ("%032x", (HEX))
> -#define print_spew_hex32(HEX)    printk_spew   ("%032x", (HEX))
> +#define print_emerg_hex32(HEX)   printk_emerg  ("%08x", (HEX))
> +#define print_alert_hex32(HEX)   printk_alert  ("%08x", (HEX))
> +#define print_crit_hex32(HEX)    printk_crit   ("%08x", (HEX))
> +#define print_err_hex32(HEX)     printk_err    ("%08x", (HEX))
> +#define print_warning_hex32(HEX) printk_warning("%08x", (HEX))
> +#define print_notice_hex32(HEX)  printk_notice ("%08x", (HEX))
> +#define print_info_hex32(HEX)    printk_info   ("%08x", (HEX))
> +#define print_debug_hex32(HEX)   printk_debug  ("%08x", (HEX))
> +#define print_spew_hex32(HEX)    printk_spew   ("%08x", (HEX))

This bit is reasonable.  I thought I had already committed it but...
I typoed and did bit widths instead of nibble widths.
> diff -uNr ./freebios2/src/northbridge/amd/amdk8/raminit.c
> ../freebios2/src/northbridge/amd/amdk8/raminit.c
>
> --- ./freebios2/src/northbridge/amd/amdk8/raminit.c 2003-08-04
> 17:59:30.000000000 -0400
>
> +++ ../freebios2/src/northbridge/amd/amdk8/raminit.c 2003-08-04
> 18:47:31.000000000 -0400
>
> @@ -1124,6 +1124,23 @@
>    }
>  }
>
> +//BY LYH //Fill next base reg with right value
> +static void fill_last(unsigned long node_id,unsigned long base)
> +{
>    unsigned i;
>    unsigned base_reg;
>    base &=0xffff0000;
>    device_t device;
> for(device = PCI_DEV(0, 0x18, 1); device <= PCI_DEV(0, 0x1f, 1); device
>
> ++= PCI_DEV(0, 1, 0)) {
>            for(i=node_id+1;i<=7;i++) {
>                    base_reg=0x40+(i<<3);
>                    pci_write_config32(device,base_reg,base);
>            }
>    }
>
> +}
> +//BY LYH END
>
> static void route_dram_accesses(const struct mem_controller *ctrl,
>    unsigned long base_k, unsigned long limit_k)
>  {
> @@ -1276,6 +1293,9 @@
>    print_debug("\r\n");
>  #endif
>    route_dram_accesses(ctrl, base_k, tom_k);
> +//BY LYH
>    fill_last(ctrl->node_id, tom_k<<2);
>
> +//BY LYH END
>    set_top_mem(tom_k);
>  }

This is a bug fix for a buggy kernel.   It is extra code but it should
not be a real problem.
> diff -uNr ./freebios2/src/southbridge/amd/amd8111/config.lb
> ../freebios2/src/southbridge/amd/amd8111/config.lb
>
> --- ./freebios2/src/southbridge/amd/amd8111/config.lb 2003-06-17
> 12:51:05.000000000 -0400
>
> +++ ../freebios2/src/southbridge/amd/amd8111/config.lb 1969-12-31
> 19:00:00.000000000 -0500
>
> @@ -1,4 +0,0 @@
> -driver amd8111_usb.o
> -driver amd8111_lpc.o
> -driver amd8111_ide.o
> -driver amd8111_acpi.o

Hmm.  Removing all of those driver statements looks wrong...

Eric

_______________________________________________
Linuxbios mailing list
[e-mail]
http://www.clustermatic.org/mailman/listinfo/linuxbios

Yh Lu's patches

YhLu  writes:
> Eric,
>
> As for cpufixup.c, actually I grab the fragment from raminit.c's
> set_top_ram.
> I don't want to make my own, because I can not access the most update code
> because of 24 hours delay.

Not a problem with what you did.  Just a problem with inclusion of the code.
The only real problem was the one line where I shifted by 12 instead of 10.
Otherwise the original is fine.
> Fill_last:
> Can someone raise the bug to LinuxKernel developer team?

Done.  Hmm.  The Linux kernel developer team is an interesting concept...
> There are Config.lb and config.lb in the southbridge/amd/amd8111.
> We should get rid of config.lb.

Agreed.  ... That is why you deleted all of the driver statements.
That makes sense now.

Eric

Yh Lu's patches

[ron minnich]

On Mon, 4 Aug 2003, YhLu wrote:
> There are Config.lb and config.lb in the southbridge/amd/amd8111.
> We should get rid of config.lb.

done and comitted.

ron