SeaBIOS

seabios@seabios.org

4 participants
3800 discussions

[PATCH 0/3] Support TPM2.0 devices on TIS host
by Stephen Douthit 27 Feb '18

27 Feb '18

Hello all, I ran into an issue getting a Infineon SLB9670 2.0 TPM running on an Intel Denverton platform, and traced the problem to what seems to be incorrect behavior of tis_get_tpm_version(). That function is returning the version of the interface, and it seems the device version is what is needed. The version check implemented here is based on the tpm2_probe() function in the Linux tpm_tis driver. This patch set applies on top of Marc-André and Stefan's "Add CRB TPM device" patch series currently under review I've tested this on real hardware with a SLB9670 hanging off of an Intel C3758, but have not regression tested with TPM 1.2 devices. Regards, Steve Stephen Douthit (3): tpm: Refactor duplicated wait code in tis_wait_sts() & crb_wait_reg() tpm: Wait for interface startup when probing tpm: Support 2.0 TPM devices connected to a TIS host src/hw/tpm_drivers.c | 114 +++++++++++++++++++++++++++------------------------ src/std/tcg.h | 1 + src/tcgbios.c | 2 +- src/tcgbios.h | 4 ++ 4 files changed, 66 insertions(+), 55 deletions(-) -- 2.14.3

2 15

reboot fails with QEMU "isapc" machine type
by Daniel P. Berrangé 27 Feb '18

27 Feb '18

I was experimenting with old MS-DOS 6.22 under QEMU with the 'isapc' machine type, and found it is not able to reboot the guest. Bisecting QEMU blamed the QEMU change that rebased from "rel-1.9.3", to "rel-1.10.0". Further bisecting SeaBIOS blames this change: commit b837e68d5a6c1a5945513f1995875445a1594c8a (refs/bisect/bad) Author: Kevin O'Connor <kevin(a)koconnor.net> Date: Mon Nov 9 15:00:19 2015 -0500 resume: Make KVM soft reboot loop detection more flexible Move the check for soft reboot loops from resume.c to shadow.c and directly check for the case where the copy of the BIOS in flash appears to be a memory alias instead. This prevents a hang if an external reboot request occurs during the BIOS memcpy. Signed-off-by: Kevin O'Connor <kevin(a)koconnor.net> I was testing as follows: $ qemu-system-x86_64 -machine isapc -monitor stdio demo.img where 'demo.img' is a disk image with MS-DOS installed (FreeDOS shows the same problem too. Not checked any other OS). Once C:\ command prompt is showing I enter (qemu) sendkey ctrl-alt-delete in the QEMU monitor console. Nothing visible will happen - the cursor will stay flashing, but it no longer responds to input. Prior to this SeaBIOS commit it will correctly reboot. Interesting side note is that this only happens with QEMUs 'isapc' machine type. The 'pc' machine type is still rebooting fine. If I enable the isa-debugcon device in QEMU I see this when the 'isapc' machine tries to reboot: In resume (status=0) In 32bit resume Attempting a hard reboot Unable to hard-reboot machine - attempting shutdown. Regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|

2 5

[PATCH v2 0/4] Add CRB TPM device
by Stefan Berger 27 Feb '18

27 Feb '18

The following series implements a TPM CRB driver. The TIS device with a TPM 2.0 seems to be ignored by Windows 10, so I implemented a simple CRB device that has now been integrated into QEMU. With the CRB device, Windows 10 correctly recognizes and exchanges packets with a TPM 2.0. Marc-André (and Stefan) --- v1->v2: - followed Kevin's suggestions for patch 4/4. Marc-André Lureau (4): x86: add readq() tpm: generalize init_timeout() tpm: use get_tpm_version() callback tpm: add TPM CRB device support src/hw/tpm_drivers.c | 224 ++++++++++++++++++++++++++++++++++++++++++++++++--- src/hw/tpm_drivers.h | 26 ++++++ src/x86.h | 5 ++ 3 files changed, 243 insertions(+), 12 deletions(-) -- 2.5.5

2 6

Re: [SeaBIOS] Mysterious "is VARVERIFY32INIT but used from" problem while trying to improve SeaBIOS
by Kevin O'Connor 27 Feb '18

27 Feb '18

On Sun, Feb 25, 2018 at 08:51:56PM +0300, Mike Banon wrote: > > > > Please see the developer documentation: > > > > I am afraid this documentation does not answer why this build problem > does not happen when I apply this patch to the standalone cloned seabios > while it does happens when I apply it to the seabios built by a coreboot The default seabios configuration is to build for qemu and not coreboot. So, by default it would not include that code and would not raise a compile time error. If you configured seabios to build for coreboot and it didn't raise a compile time error then that seems like a build defect. -Kevin

1 0

[PATCH 1/2] floppy: Introduce the floppy_clear_irq_flag() function
by Nikolay Nikolov 26 Feb '18

26 Feb '18

Signed-off-by: Nikolay Nikolov <nickysn(a)users.sourceforge.net> --- src/hw/floppy.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) diff --git a/src/hw/floppy.c b/src/hw/floppy.c index 9e6647d..1120d63 100644 --- a/src/hw/floppy.c +++ b/src/hw/floppy.c @@ -220,11 +220,18 @@ floppy_disable_controller(void) floppy_dor_mask(FLOPPY_DOR_IRQ | FLOPPY_DOR_RESET, 0); } -static int -floppy_wait_irq(void) +static inline void +floppy_clear_irq_flag(void) { u8 frs = GET_BDA(floppy_recalibration_status); SET_BDA(floppy_recalibration_status, frs & ~FRS_IRQ); +} + +static int +floppy_wait_irq(void) +{ + floppy_clear_irq_flag(); + u8 frs; u32 end = timer_calc(FLOPPY_IRQ_TIMEOUT); for (;;) { if (timer_check(end)) { -- 2.14.3

3 4

[PATCH v2 0/4] Add CRB TPM device
by Stefan Berger 26 Feb '18

26 Feb '18

The following series implements a TPM CRB driver. The TIS device with a TPM 2.0 seems to be ignored by Windows 10, so I implemented a simple CRB device that has now been integrated into QEMU. With the CRB device, Windows 10 correctly recognizes and exchanges packets with a TPM 2.0. Marc-André (and Stefan) Marc-André Lureau (4): x86: add readq() tpm: generalize init_timeout() tpm: use get_tpm_version() callback tpm: add TPM CRB device support src/hw/tpm_drivers.c | 226 ++++++++++++++++++++++++++++++++++++++++++++++++--- src/hw/tpm_drivers.h | 26 ++++++ src/x86.h | 5 ++ 3 files changed, 245 insertions(+), 12 deletions(-) -- 2.5.5

3 7

Re: [SeaBIOS] Mysterious "is VARVERIFY32INIT but used from" problem while trying to improve SeaBIOS
by Kevin O'Connor 25 Feb '18

25 Feb '18

On Sun, Feb 25, 2018 at 03:05:06AM +0300, Mike Banon wrote: > Friends, I need your help. While trying to improve SeaBIOS I got stuck > at this very strange problem - which does not appear when you do a > standalone build of SeaBIOS but it happens when you try building a > coreboot together with SeaBIOS. Steps to reproduce: Please see the developer documentation: https://www.seabios.org/Developer_Documentation In particular, read the "memory model" and "execution and code phase" documents. -Kevin

2 1

Re: [SeaBIOS] Mysterious "is VARVERIFY32INIT but used from" problem while trying to improve SeaBIOS
by Paul Menzel 25 Feb '18

25 Feb '18

Dear Mike, Am 25.02.2018 um 01:05 schrieb Mike Banon: > Friends, I need your help. While trying to improve SeaBIOS I got stuck > at this very strange problem - which does not appear when you do a > standalone build of SeaBIOS but it happens when you try building a > coreboot together with SeaBIOS. Steps to reproduce: Please attach your patch, so that it can easily applied and the build problem reproduced. […] Kind regards, Paul

2 1

Mysterious "is VARVERIFY32INIT but used from" problem while trying to improve SeaBIOS
by Mike Banon 25 Feb '18

25 Feb '18

Friends, I need your help. While trying to improve SeaBIOS I got stuck at this very strange problem - which does not appear when you do a standalone build of SeaBIOS but it happens when you try building a coreboot together with SeaBIOS. Steps to reproduce: add this innocent looking function to the beginning of ./coreboot/payloads/external/SeaBIOS/seabios/src/fw/coreboot.c : int // doesn't matter if static or not testfunc(void) { int *c = malloc_tmp(sizeof(int)); // doesn't matter if _tmp or _tmphigh // free(c); // <--- commented out because doesn't affect the results return 0; } and try calling it from "cbfs_copyfile" function of the same file: static int cbfs_copyfile(struct romfile_s *file, void *dst, u32 maxlen) { ... void *temp = malloc_tmphigh(size); ... int ret = ulzma(dst, maxlen, temp, size); testfunc(); // <--- added ... } Then I get this problem while building a coreboot with together SeaBIOS : ... Compiling whole program out/code32seg.o Compiling whole program out/ccode16.o Compiling to assembler out/src/asm-offsets.s Generating offset file out/asm-offsets.h Compiling (16bit) out/romlayout.o Building ld scripts ERROR: .data.varinit.. is VARVERIFY32INIT but used from ['.text.runtime..', '.text.do_boot', '.text.malloc_tmp'] Makefile:162: recipe for target 'out/romlayout16.lds' failed make[2]: *** [out/romlayout16.lds] Error 1 Makefile:80: recipe for target 'build' failed make[1]: *** [build] Error 2 payloads/external/Makefile.inc:73: recipe for target 'payloads/external/SeaBIOS/seabios/out/bios.bin.elf' failed make: *** [payloads/external/SeaBIOS/seabios/out/bios.bin.elf] Error 2 However, if I try adding this function to the end of "coreboot_cbfs_init" function I don't get such an error: void coreboot_cbfs_init(void) { ... process_links_file(); testfunc(); // <--- works OK there } What is a difference between "cbfs_copyfile" and "coreboot_cbfs_init" that affects this behaviour? And why it is OK to call "malloc_" from inside the "cbfs_copyfile" function --- void *temp = malloc_tmphigh(size); --- but we couldn't call there our testfunc() which calls "malloc_" without running into this problem? Best regards, Mike Banon

1 0

Bug report for 'vga none' qemu option not working any more
by Sergey Aleynikov 23 Feb '18

23 Feb '18

I'm running a lot of qemu + kvm + seabios VMs and after upgrade to Debian 9 I found that some of them stopped booting. After trying to change kernel/qemu/seabios versions, I've nailed down the cause to the following: # VM definition qemu-system-x86_64 \ -enable-kvm \ -machine q35 \ -m 512 \ -nodefaults \ -cpu host \ -smp 1,sockets=1,cores=1,threads=1 \ -bios /opt/seabios/out/bios.bin \ -device virtio-scsi-pci,id=scsi \ -drive file="disk.img",id=disk,format=raw,if=none,cache=unsafe,aio=threads \ -device scsi-hd,drive=disk,bus=scsi.0,removable=off \ -device virtio-net-pci,netdev=net0,mac=$("../qemu-mac-hasher.py" "test") \ -netdev tap,id=net0,script="../ifup.sh",downscript="../ifdown.sh" \ -chardev stdio,id=seabios -device isa-debugcon,iobase=0x402,chardev=seabios \ -vga none \ -nographic This VM doesn't boot with modern seabios, but it used to work. And now, if I change 'vga none' to 'vga std' in it, it starts booting again. I've performed git bisect on the seabios repository and found the following commit to break the old setup: 1d9e87b937d646be1950695f9ead35100d5ebbe6 is the first bad commit commit 1d9e87b937d646be1950695f9ead35100d5ebbe6 Author: Gerd Hoffmann <kraxel(a)redhat.com> Date: Fri Jun 26 09:44:00 2015 +0200 virtio: run drivers in 32bit mode virtio version 1.0 registers can (and actually do in the qemu implementation) live in mmio space. So we must run the blk and scsi virtio drivers in 32bit mode, otherwise we can't access them. This also allows to drop a bunch of GET_LOWFLAT calls from the virtio code in the following patches. Signed-off-by: Gerd Hoffmann <kraxel(a)redhat.com> :100644 100644 f97b1bd9f91df7417e7bd5a1ebd7182098a43296 e287530d0008e1d876a0bb9f9ff8a43266d7e1bd M Makefile :040000 040000 70b419ced4f58a39143c92897097a08b5548cbf4 93f7d4d939b5e0a14102a3cb718b1be5a2f12e60 M src Here're links to a full successful boot log (with 'vga std') and some portion of a failing boot log (as it just loops eternally): https://gist.github.com/dur-randir/80b7d9a84cbdb8591ec05abe2fc62604 https://gist.github.com/dur-randir/52c31d2abe3150759abd3e8ca140959a Those logs were obtained from seabios 1d9e87b937 - the first broken revision, but I can provide them for master, as it's still broken. Best regards, Sergey Aleyinkov

1 0

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

SeaBIOS