SeaBIOS

seabios@seabios.org

5 participants
3802 discussions

Re: [SeaBIOS] [PATCH] smbios: generate the correct the number of SMBIOS type 4 reocrds
by Kevin O'Connor 11 Dec '13

11 Dec '13

On Wed, Dec 11, 2013 at 12:05:34PM -0600, Bret Ketchum wrote: > Thanks. What time-frame are we talking about; looks like qemu 1.7.0 is > about to be released with seabios-1.7.3 and includes ACPI table generation. > I don't see SMBIOS table/record generation support however. > Most smbios tables can already be passed from QEMU to SeaBIOS. It would be great to expand that support for any remaining tables/fields. I don't know of a particular time frame to do that though. -Kevin

1 0

Re: [SeaBIOS] [coreboot] Patch number 4 for floppy in SeaBIOS
by Gelip 11 Dec '13

11 Dec '13

In serial log is: SeaBIOS (version rel-1.7.3-123-gfeb0284 I build coreboot like this: cd /coreboot/build rm coreboot.rom rm -R seabios git clone https://github.com/KevinOConnor/seabios -b testing cd seabios make menuconfig -> coreboot and debug 8 patch -p1 < fdd.p patch -p1 < config.p make cd ../.. make menuconfig and Payload None make ./build/cbfstool build/coreboot.rom add-payload -f build/seabios/out/bios.bin.elf -n fallback/payload -c lzma ./build/seabios/scripts/encodeint.py floppy0 4 ./build/cbfstool build/coreboot.rom add -f floppy0 -n etc/floppy0 -t raw ./build/seabios/scripts/readserial.py /dev/ttyS0 115200

1 0

Re: [SeaBIOS] [coreboot] Patch number 4 for floppy in SeaBIOS
by Gelip 11 Dec '13

11 Dec '13

10.076: Booting from Floppy... 10.076: enter handle_13: 10.076: a=00000201 b=00000000 c=00000001 d=00000000 ds=0000 es=07c0 ss=df80 10.076: si=00000000 di=00000000 bp=00000000 sp=0000f910 cs=f000 ip=cb8b f=0202 10.076: disk_op d=0x000f5500 lba=0 buf=0x00007c00 count=1 cmd=2 10.076: Floppy_drive_recal 0 10.076: Floppy_enable_controller 10.076: handle_0e 10.076: handle_0e 10.430: handle_0e 10.430: Floppy_media_sense on drive 0 found rate 0 10.622: handle_0e 10.622: Booting from 0000:7c00 10.622: enter handle_13: 10.622: a=00000201 b=00000700 c=00000002 d=00000100 ds=0000 es=0000 ss=df80 10.622: si=00007bd2 di=00000700 bp=00007c00 sp=0000f910 cs=0000 ip=7d6e f=0002 10.622: disk_op d=0x000f5500 lba=19 buf=0x00000700 count=1 cmd=2 10.811: handle_0e 10.811: enter handle_13: 10.811: a=00000201 b=00000700 c=00000010 d=00000100 ds=0000 es=0000 ss=df80 10.811: si=00007bd2 di=00000002 bp=00007c00 sp=0000f910 cs=0000 ip=7d6e f=0002 10.811: disk_op d=0x000f5500 lba=33 buf=0x00000700 count=1 cmd=2 10.817: Floppy seek 0 to cyl 0 10.817: handle_0e 10.941: handle_0e 10.941: enter handle_13: 10.941: a=00000201 b=00000900 c=00000011 d=00000100 ds=0000 es=0000 ss=df80 10.941: si=00007bd2 di=00000002 bp=00007c00 sp=0000f910 cs=0000 ip=7d6e f=0006 10.944: disk_op d=0x000f5500 lba=34 buf=0x00000900 count=1 cmd=2 10.944: Floppy seek 0 to cyl 0 10.944: handle_0e 11.129: handle_0e 11.129: enter handle_13: 11.129: a=00000201 b=00000b00 c=00000012 d=00000100 ds=0000 es=0000 ss=df80 11.129: si=00007bd2 di=00000002 bp=00007c00 sp=0000f910 cs=0000 ip=7d6e f=0006 11.131: disk_op d=0x000f5500 lba=35 buf=0x00000b00 count=1 cmd=2 11.131: Floppy seek 0 to cyl 0 11.131: handle_0e 11.315: handle_0e 11.315: enter handle_13: 11.315: a=00000201 b=00000d00 c=00000101 d=00000000 ds=0000 es=0000 ss=df80 11.315: si=00007bd2 di=00000002 bp=00007c00 sp=0000f910 cs=0000 ip=7d6e f=0002 11.318: disk_op d=0x000f5500 lba=36 buf=0x00000d00 count=1 cmd=2 11.697: handle_0e 11.697: floppy error: 40 04 10 01 00 01 02 11.697: invalid basic_access:96: 11.697: a=00000200 b=00000d00 c=00000101 d=00000000 ds=0000 es=0000 ss=df80 11.697: si=00007bd2 di=00000002 bp=00007c00 sp=0000f910 cs=0000 ip=7d6e f=0002

2 2

Re: [SeaBIOS] [PATCH] smbios: generate the correct the number of SMBIOS type 4 reocrds
by Kevin O'Connor 11 Dec '13

11 Dec '13

On Wed, Dec 11, 2013 at 08:13:36AM -0600, Bret Ketchum wrote: > This patch uses cpuid to determine the number of packages/sockets in which > to > generate the proper number of DMI/SMBIOS Type 4 (Processor Information) > records. > > Use "dmidevcode -t 4" to show that only the configured number of > packages/sockets are shown versus the number of CPUs > (sockets*cores*threads). Thanks Bret. The SMBIOS stored in seabios is really a function of QEMU. So, discussions on its correct content and changes to it need to be discussed on the QEMU mailing list. Going forward, we want to move all the ACPI, SMBIOS, and MPTable changes to the QEMU side. -Kevin

2 1

Re: [SeaBIOS] [coreboot] Patch number 4 for floppy in SeaBIOS
by Gelip 11 Dec '13

11 Dec '13

"If you weren't using 1.44Mb floppies, then something went very wrong. Otherwise, I'm not sure how the above could be related to the failure you are seeing." I always use floppy 3.5' 1.44MB. Does the use of FDD depend on CMOS Floppy Disk Controller settings? I noticed that WinXP display FDD in My Computer window (http://savepic.net/4018623.htm): - if coreboot + CMOS default from vendor BIOS: A: 3.5' - if coreboot + CMOS clear: A: 5.25' and B: 5.25' Maybe need apply CMOS.layout file to coreboot to resolve this problem? I just do not know how to make this file for Abit BF6 mobo.

2 1

Re: [SeaBIOS] [coreboot] Patch number 4 for floppy in SeaBIOS
by Gelip 11 Dec '13

11 Dec '13

08.142: Booting from Floppy... 08.142: enter handle_13: 08.143: a=00000201 b=00000000 c=00000001 d=00000000 ds=0000 es=07c0 ss=df80 08.142: si=00000000 di=00000000 bp=00000000 sp=0000f9b0 cs=f000 ip=cbdf f=0202 08.143: disk_op d=0x000f55a0 lba=0 buf=0x00007c00 count=1 cmd=2 08.143: Floppy_drive_recal 0 08.143: Floppy_enable_controller 08.146: handle_0e 08.145: handle_0e 08.530: handle_0e 08.530: Floppy send specify 4f 10 08.530: Floppy_media_sense on drive 0 found rate 0 08.722: handle_0e 08.722: Booting from 0000:7c00 08.722: enter handle_13: 08.722: a=00000201 b=00000700 c=00000002 d=00000100 ds=0000 es=0000 ss=df80 08.723: si=00007bd2 di=00000700 bp=00007c00 sp=0000f9b0 cs=0000 ip=7d6e f=0002 08.722: disk_op d=0x000f55a0 lba=19 buf=0x00000700 count=1 cmd=2 08.915: handle_0e 08.915: enter handle_13: 08.915: a=00000201 b=00000700 c=00000010 d=00000100 ds=0000 es=0000 ss=df80 08.915: si=00007bd2 di=00000002 bp=00007c00 sp=0000f9b0 cs=0000 ip=7d6e f=0002 08.916: disk_op d=0x000f55a0 lba=33 buf=0x00000700 count=1 cmd=2 09.045: handle_0e 09.045: enter handle_13: 09.045: a=00000201 b=00000900 c=00000011 d=00000100 ds=0000 es=0000 ss=df80 09.046: si=00007bd2 di=00000002 bp=00007c00 sp=0000f9b0 cs=0000 ip=7d6e f=0006 09.046: disk_op d=0x000f55a0 lba=34 buf=0x00000900 count=1 cmd=2 09.240: handle_0e 09.240: enter handle_13: 09.240: a=00000201 b=00000b00 c=00000012 d=00000100 ds=0000 es=0000 ss=df80 09.240: si=00007bd2 di=00000002 bp=00007c00 sp=0000f9b0 cs=0000 ip=7d6e f=0006 09.240: disk_op d=0x000f55a0 lba=35 buf=0x00000b00 count=1 cmd=2 09.428: handle_0e 09.428: enter handle_13: 09.428: a=00000201 b=00000d00 c=00000101 d=00000000 ds=0000 es=0000 ss=df80 09.428: si=00007bd2 di=00000002 bp=00007c00 sp=0000f9b0 cs=0000 ip=7d6e f=0002 09.428: disk_op d=0x000f55a0 lba=36 buf=0x00000d00 count=1 cmd=2 09.810: handle_0e 09.810: floppy error: 40 04 10 01 00 01 02 09.810: invalid basic_access:96: 09.810: a=00000200 b=00000d00 c=00000101 d=00000000 ds=0000 es=0000 ss=df80 09.810: si=00007bd2 di=00000002 bp=00007c00 sp=0000f9b0 cs=0000 ip=7d6e f=0002

2 1

[PATCH] smbios: generate the correct the number of SMBIOS type 4 reocrds
by Bret Ketchum 11 Dec '13

11 Dec '13

This patch uses cpuid to determine the number of packages/sockets in which to generate the proper number of DMI/SMBIOS Type 4 (Processor Information) records. Use "dmidevcode -t 4" to show that only the configured number of packages/sockets are shown versus the number of CPUs (sockets*cores*threads). Signed-off-by: Bret Ketchum <bcketchum at gmail.com> --- src/fw/smbios.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 48 insertions(+), 2 deletions(-) diff --git a/src/fw/smbios.c b/src/fw/smbios.c index affb9be..23869b1 100644 --- a/src/fw/smbios.c +++ b/src/fw/smbios.c @@ -505,6 +505,52 @@ smbios_init_type_127(void *start) #define TEMPSMBIOSSIZE (32 * 1024) +static inline void +cpuid_count(u32 index, u32 sub, u32 *eax, u32 *ebx, u32 *ecx, u32 *edx) +{ + asm volatile("cpuid" + : "=a" (*eax), "=b" (*ebx), "=c" (*ecx), "=d" (*edx) + : "0" (index), "c" (sub)); +} + +static int package_count(void) +{ + char vendor[12]; + int ret = CountCPUs; + u32 cpuid_features, cpuid_high, cpuid_procinfo, cpuid_signature; + u32 eax, ebx, ecx, edx; + + cpuid(0, &cpuid_high, &ebx, &ecx, &edx); + ((u32 *)&vendor)[0] = ebx; + ((u32 *)&vendor)[1] = edx; + ((u32 *)&vendor)[2] = ecx; + + cpuid(1, &cpuid_signature, &cpuid_procinfo, &ecx, &cpuid_features); + + if (memcmp(vendor, "GenuineIntel", sizeof(vendor)) == 0) { + int logical = (cpuid_features & 0x10000000) != 0 + ? (cpuid_procinfo >> 16) & 0xff + : 1; + + cpuid_count(4, 0, &eax, &ebx, &ecx, &edx); + + /* number of packages/sockets equals the number ACPI IDs + * divided by the "Maximum number of addressable IDs for + * logical processors in this physical package. + */ + ret = MaxCountCPUs / ((logical == 1) + ? (((eax >> 26) & 0x3f) + 1) + : logical); + + } else { + cpuid(0x80000008, &eax, &ebx, &ecx, &edx); + + ret = ((cpuid_procinfo >> 16) & 0xff) / ((ecx & 0xff) + 1); + } + + return(ret); +} + void smbios_setup(void) { @@ -538,8 +584,8 @@ smbios_setup(void) add_struct(1, p); add_struct(3, p); - int cpu_num; - for (cpu_num = 1; cpu_num <= MaxCountCPUs; cpu_num++) + int cpu_num, pkg_count = package_count(); + for (cpu_num = 1; cpu_num <= pkg_count; cpu_num++) add_struct(4, p, cpu_num); int ram_mb = (RamSize + RamSizeOver4G) >> 20; -- 1.8.3.1

1 0

DMI BIOS date & System BIOS date not the same in AIDA64 on WinXP
by Gelip 11 Dec '13

11 Dec '13

coreboot 4.0-4772 + SeaBIOS 1.7.3-117 I noticed that BIOS date in AID64 on WinXP is different in Computer > DMI and Motherboard > BIOS Computer > DMI > BIOS > Release Date: 12/11/2013 (this is date build coreboot) Motherboard > BIOS > System BIOS Date: 06/23/99 (this is probably SeaBIOS date from src/misc.c file) In vendor bios, date is the same in both locations. DMI date: http://savepic.net/4009104.htm System BIOS date: http://savepic.net/4058259.htm

1 0

Re: [SeaBIOS] SMBIOS Type 4
by Paul Menzel 11 Dec '13

11 Dec '13

Dear Bret, thanks a lot for your patch! Am Montag, den 09.12.2013, 11:54 -0600 schrieb Bret Ketchum: > Here's a patch which uses cpuid to determine the number of packages and > build the correct number of type 4s: […] Could you please send a `git format-patch` formatted patch with a commit message (including summary and body)? That would be awesome. How to test your patch? For example with `dmidecode`? Thanks, Paul

1 0

Re: [SeaBIOS] SMBIOS Type 4
by Bret Ketchum 09 Dec '13

09 Dec '13

Here's a patch which uses cpuid to determine the number of packages and build the correct number of type 4s: diff --git a/src/fw/smbios.c b/src/fw/smbios.c index affb9be..23869b1 100644 --- a/src/fw/smbios.c +++ b/src/fw/smbios.c @@ -505,6 +505,52 @@ smbios_init_type_127(void *start) #define TEMPSMBIOSSIZE (32 * 1024) +static inline void +cpuid_count(u32 index, u32 sub, u32 *eax, u32 *ebx, u32 *ecx, u32 *edx) +{ + asm volatile("cpuid" + : "=a" (*eax), "=b" (*ebx), "=c" (*ecx), "=d" (*edx) + : "0" (index), "c" (sub)); +} + +static int package_count(void) +{ + char vendor[12]; + int ret = CountCPUs; + u32 cpuid_features, cpuid_high, cpuid_procinfo, cpuid_signature; + u32 eax, ebx, ecx, edx; + + cpuid(0, &cpuid_high, &ebx, &ecx, &edx); + ((u32 *)&vendor)[0] = ebx; + ((u32 *)&vendor)[1] = edx; + ((u32 *)&vendor)[2] = ecx; + + cpuid(1, &cpuid_signature, &cpuid_procinfo, &ecx, &cpuid_features); + + if (memcmp(vendor, "GenuineIntel", sizeof(vendor)) == 0) { + int logical = (cpuid_features & 0x10000000) != 0 + ? (cpuid_procinfo >> 16) & 0xff + : 1; + + cpuid_count(4, 0, &eax, &ebx, &ecx, &edx); + + /* number of packages/sockets equals the number ACPI IDs + * divided by the "Maximum number of addressable IDs for + * logical processors in this physical package. + */ + ret = MaxCountCPUs / ((logical == 1) + ? (((eax >> 26) & 0x3f) + 1) + : logical); + + } else { + cpuid(0x80000008, &eax, &ebx, &ecx, &edx); + + ret = ((cpuid_procinfo >> 16) & 0xff) / ((ecx & 0xff) + 1); + } + + return(ret); +} + void smbios_setup(void) { @@ -538,8 +584,8 @@ smbios_setup(void) add_struct(1, p); add_struct(3, p); - int cpu_num; - for (cpu_num = 1; cpu_num <= MaxCountCPUs; cpu_num++) + int cpu_num, pkg_count = package_count(); + for (cpu_num = 1; cpu_num <= pkg_count; cpu_num++) add_struct(4, p, cpu_num); int ram_mb = (RamSize + RamSizeOver4G) >> 20; On Fri, Dec 6, 2013 at 1:22 PM, Bret Ketchum <bcketchum(a)gmail.com> wrote: > > Looking at the 1.7.3 source it appears that smbios_setup creates a > type 4 entry for each core rather than one per package/socket as would be > typical of a bare metal bios. > > Is this behavior by design? Thanks in advance.... > > Bret >

1 0

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

SeaBIOS