SeaBIOS July 2015

seabios@seabios.org

14 participants
33 discussions

[PATCH 0/7] kvmtool support for seabios
by Gerd Hoffmann 07 Jul '15

07 Jul '15

Hi, Played around a bit with kvmtool. Got annonyed because you can't just pass it a guest image to boot from. Decided to check what it takes to run seabios in kvmtool, so it can load the guest kernel from the disk image. So, here are the results ... There isn't much work involved at the seabios side. Add a new config option. Tweak kconfig. Some basic windup on the code. Loads kernel from disk. Yea! Success! Well. That only uncovers a bunch of issues in kvmtool ... (1) There seems to be no way to figure the amout of memory available. For now I've hardcoded 128 MB in seabios. See patch #3. (2) Something corrupts virtio rings. Not clear yet whenever seabios or kvmtool is at fault here. See patch #7. (3) kvmtool doesn't emulate a pci host bridge. Linux kernel is unhappy and throws an error on pci initialization. kvmtool tackles that by appending "pci=conf1" to the command line when booting a kernel directly. Which of course doesn't fly if you wanna boot the fedora cloud image downloaded from the internet as-is. Booting the image once in qemu to tweak the boot loader config gets over it. (4) Linux kernel fails on virtio-blk initialization. Bummer. /me suspects kvmtool doesn't implement proper virtio device reset, so the device handover from seabios to linux kernel doesn't work. So, the kernel boots fine but fails to mount the root filesystem. After a while dracut decides to not wait any longer for the root filesystem to show up and drops into the emergency shell, where you can play around a bit. enjoy, Gerd Patches also available in the git repository at: git://git.kraxel.org/seabios kvmtool ---------------------------------------------------------------- Gerd Hoffmann (7): kvmtool: initial support kvmtool: tweak kconfig for emulated hardware kvmtool: add (fake) ram detection kvmtool: detect pci devices kvmtool: uses mmio for legacy bar 0 kvmtool: no keyboard, don't need boot menu kvmtool: support larger virtio queues Makefile | 1 + src/Kconfig | 24 ++++++++++++++++++++++-- src/fw/paravirt.c | 20 ++++++++++++++++++++ src/fw/paravirt.h | 3 +++ src/hw/virtio-pci.c | 11 ++++++++--- src/hw/virtio-ring.h | 2 +- src/post.c | 4 +++- 7 files changed, 58 insertions(+), 7 deletions(-)

1 7

Re: [SeaBIOS] SeaBIOS stuck at timer setup
by Patrick Rudolph 06 Jul '15

06 Jul '15

Am 2015-07-06 15:25, schrieb Kevin O'Connor: > On Sun, Jul 05, 2015 at 07:30:29PM +0200, Patrick Rudolph wrote: >> Hello, >> I'm trying to use SeaBIOS, but it's stuck at timer init. >> I've located the problem at line 71 in file src/hw/timer.c. >> It tries to write IO register 0x43 and freezes. >> >> Do you have any idea what could be wrong ? >> >> Test system: >> * Intel Pentium CPU G2130 >> * Gigabyte GA-B75M-D3H >> * SeaBIOS 1.7.5 and SeaBIOS git f24eb2f853d4aa28814761e0bbc7df78149f8029 >> * coreboot default configuration > > Can you send the full debug log of the failed boot? > > A failed write to 0x43 sounds very odd - it could be an incomplete > hardware configuration in coreboot, or it could be the legacy timer is > not fully supported on that board (though that seems unlikely). You > could comment out the call to timer_setup() in post.c and see if the > boot progresses further. > > -Kevin Hi Kevin, I've attached the console output. I commented out this code, but other hardware initialization fails, too. The clock, xhci, ehci, ps2port are not working. It looks like a generic problem related to inb or outb ! Regards, Patrick

2 1

SeaBIOS stuck at timer setup
by Patrick Rudolph 06 Jul '15

06 Jul '15

Hello, I'm trying to use SeaBIOS, but it's stuck at timer init. I've located the problem at line 71 in file src/hw/timer.c. It tries to write IO register 0x43 and freezes. Do you have any idea what could be wrong ? Test system: * Intel Pentium CPU G2130 * Gigabyte GA-B75M-D3H * SeaBIOS 1.7.5 and SeaBIOS git f24eb2f853d4aa28814761e0bbc7df78149f8029 * coreboot default configuration Kind Regards, Patrick Rudolph

2 1

[PATCH] Make sure all code checks for malloc failures
by Kevin O'Connor 04 Jul '15

04 Jul '15

This is the result of an audit of callers of the malloc_XXX() and memalign_XXX() calls. All callers need to check if these functions return NULL. Signed-off-by: Kevin O'Connor <kevin(a)koconnor.net> --- src/cdrom.c | 1 - src/hw/ahci.c | 8 ++++++++ src/hw/megasas.c | 6 +++++- src/hw/ramdisk.c | 2 +- src/hw/usb-xhci.c | 9 ++++++++- src/hw/usb.c | 4 +++- 6 files changed, 25 insertions(+), 5 deletions(-) diff --git a/src/cdrom.c b/src/cdrom.c index 7ee36d9..ba02340 100644 --- a/src/cdrom.c +++ b/src/cdrom.c @@ -123,7 +123,6 @@ cdrom_prepboot(void) struct drive_s *drive = malloc_fseg(sizeof(*drive)); if (!drive) { warn_noalloc(); - free(drive); return; } cdemu_drive_gf = drive; diff --git a/src/hw/ahci.c b/src/hw/ahci.c index 3193d81..0d71cc4 100644 --- a/src/hw/ahci.c +++ b/src/hw/ahci.c @@ -405,6 +405,14 @@ static struct ahci_port_s* ahci_port_realloc(struct ahci_port_s *port) port->list = memalign_high(1024, 1024); port->fis = memalign_high(256, 256); port->cmd = memalign_high(256, 256); + if (!port->list || !port->fis || !port->cmd) { + warn_noalloc(); + free(port->list); + free(port->fis); + free(port->cmd); + free(port); + return NULL; + } ahci_port_writel(port->ctrl, port->pnr, PORT_LST_ADDR, (u32)port->list); ahci_port_writel(port->ctrl, port->pnr, PORT_FIS_ADDR, (u32)port->fis); diff --git a/src/hw/megasas.c b/src/hw/megasas.c index b2a65e4..6677977 100644 --- a/src/hw/megasas.c +++ b/src/hw/megasas.c @@ -241,7 +241,10 @@ static void megasas_scan_target(struct pci_device *pci, u32 iobase) { struct mfi_ld_list_s ld_list; struct megasas_cmd_frame *frame = memalign_tmp(256, sizeof(*frame)); - int i; + if (!frame) { + warn_noalloc(); + return; + } memset(&ld_list, 0, sizeof(ld_list)); memset_fl(frame, 0, sizeof(*frame)); @@ -258,6 +261,7 @@ static void megasas_scan_target(struct pci_device *pci, u32 iobase) if (megasas_fire_cmd(pci->device, iobase, frame) == 0) { dprintf(2, "%d LD found\n", ld_list.count); + int i; for (i = 0; i < ld_list.count; i++) { dprintf(2, "LD %d:%d state 0x%x\n", ld_list.lds[i].target, ld_list.lds[i].lun, diff --git a/src/hw/ramdisk.c b/src/hw/ramdisk.c index 1177bc0..6b44c83 100644 --- a/src/hw/ramdisk.c +++ b/src/hw/ramdisk.c @@ -7,7 +7,7 @@ #include "biosvar.h" // GET_GLOBALFLAT #include "block.h" // struct drive_s #include "bregs.h" // struct bregs -#include "malloc.h" // malloc_fseg +#include "malloc.h" // memalign_tmphigh #include "memmap.h" // add_e820 #include "output.h" // dprintf #include "romfile.h" // romfile_findprefix diff --git a/src/hw/usb-xhci.c b/src/hw/usb-xhci.c index fd58334..41a6a3f 100644 --- a/src/hw/usb-xhci.c +++ b/src/hw/usb-xhci.c @@ -921,8 +921,14 @@ xhci_alloc_pipe(struct usbdevice_s *usbdev usb_desc2pipe(&pipe->pipe, usbdev, epdesc); pipe->epid = epid; pipe->reqs.cs = 1; - if (eptype == USB_ENDPOINT_XFER_INT) + if (eptype == USB_ENDPOINT_XFER_INT) { pipe->buf = malloc_high(pipe->pipe.maxpacket); + if (!pipe->buf) { + warn_noalloc(); + free(pipe); + return NULL; + } + } // Allocate input context and initialize endpoint info. struct xhci_inctx *in = xhci_alloc_inctx(usbdev, epid); @@ -988,6 +994,7 @@ xhci_alloc_pipe(struct usbdevice_s *usbdev return &pipe->pipe; fail: + free(pipe->buf); free(pipe); free(in); return NULL; diff --git a/src/hw/usb.c b/src/hw/usb.c index 1b4ea8b..2d5c224 100644 --- a/src/hw/usb.c +++ b/src/hw/usb.c @@ -249,8 +249,10 @@ get_device_config(struct usb_pipe *pipe) return NULL; void *config = malloc_tmphigh(cfg.wTotalLength); - if (!config) + if (!config) { + warn_noalloc(); return NULL; + } req.wLength = cfg.wTotalLength; ret = usb_send_default_control(pipe, &req, config); if (ret) { -- 1.9.3

1 1

[PATCH] bootorder: Update "extra pci root" buses bootorder format to match qemu
by Kevin O'Connor 04 Jul '15

04 Jul '15

The QEMU code will use the prefix "/pci@i0cf8,%x/" for devices on extra pci root buses, so change the SeaBIOS code to match that. Signed-off-by: Kevin O'Connor <kevin(a)koconnor.net> --- src/boot.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/boot.c b/src/boot.c index ec59c37..e0f73a3 100644 --- a/src/boot.c +++ b/src/boot.c @@ -112,9 +112,9 @@ build_pci_path(char *buf, int max, const char *devname, struct pci_device *pci) if (pci->parent) { p = build_pci_path(p, max, "pci-bridge", pci->parent); } else { - if (pci->rootbus) - p += snprintf(p, max, "/pci-root@%x", pci->rootbus); p += snprintf(p, buf+max-p, "%s", FW_PCI_DOMAIN); + if (pci->rootbus) + p += snprintf(p, buf+max-p, ",%x", pci->rootbus); } int dev = pci_bdf_to_dev(pci->bdf), fn = pci_bdf_to_fn(pci->bdf); -- 1.9.3

1 1

Re: [SeaBIOS] Reducing SeaBIOS kernel entry time
by Kevin O'Connor 03 Jul '15

03 Jul '15

On Fri, Jul 03, 2015 at 10:33:56AM +0100, Stefan Hajnoczi wrote: > Hi, > qboot (https://github.com/bonzini/qboot) is a stripped down firmware > providing only what is needed to boot a Linux kernel on x86. I wonder > if there is room to achieve similar results with SeaBIOS for the > lightweight virtual machine use case. > > The goal is to enter the kernel within 20 milliseconds. (That's the > number I measured with qboot.) > > Kevin: Have you looked at SeaBIOS startup time and do you think this > use case is within the scope of what SeaBIOS is intended for? > > Marc is beginning to profile boot times and any pointers would be helpful. Hi Stefan, I took a look a month or so ago: http://article.gmane.org/gmane.comp.emulators.kvm.devel/136207 Last test I ran I had SeaBIOS down to 16ms (not including the time to deploy or jump to the linux kernel). More reductions are also possible - the above was just with Kconfig settings. Let me know if you want more info on how I did that. -Kevin

2 1

Re: [SeaBIOS] Reducing SeaBIOS kernel entry time
by Peter Stuge 03 Jul '15

03 Jul '15

Stefan Hajnoczi wrote: > Hi, > qboot (https://github.com/bonzini/qboot) is a stripped down firmware > providing only what is needed to boot a Linux kernel on x86. Incredible, you have re-invented coreboot! Well, maybe not, it is Red Hat after all. How about putting your efforts into the existing project instead, that way you can also draw on the significant experience within that project. If you want to optimize for the Linux special case then you should not be using anything BIOS-related at all. Try coreboot with kernel as payload. //Peter

4 3

[PATCH 0/3] virtio: pci cfg access
by Gerd Hoffmann 03 Jul '15

03 Jul '15

Hi, As discussed this implements virtio regions access via pci cfg space window. Used in case bars are mapped above 4G. Patch #3 is for testing and not intended to be merged. cheers, Gerd Gerd Hoffmann (3): virtio: uninline _vp_{read,write} virtio: pci cfg access [testing] map 64bit bars high unconditionally. src/fw/pciinit.c | 2 +- src/hw/virtio-pci.c | 224 +++++++++++++++++++++++++++++++++++++++++++++++++--- src/hw/virtio-pci.h | 92 ++++----------------- 3 files changed, 227 insertions(+), 91 deletions(-) -- 1.8.3.1

1 3

Reducing SeaBIOS kernel entry time
by Stefan Hajnoczi 03 Jul '15

03 Jul '15

Hi, qboot (https://github.com/bonzini/qboot) is a stripped down firmware providing only what is needed to boot a Linux kernel on x86. I wonder if there is room to achieve similar results with SeaBIOS for the lightweight virtual machine use case. The goal is to enter the kernel within 20 milliseconds. (That's the number I measured with qboot.) Kevin: Have you looked at SeaBIOS startup time and do you think this use case is within the scope of what SeaBIOS is intended for? Marc is beginning to profile boot times and any pointers would be helpful. Thanks, Stefan

1 0

[PATCH v3 00/25] virtio: add version 1.0 support, move to 32bit
by Gerd Hoffmann 03 Jul '15

03 Jul '15

Hi, This patch series adds virtio 1.0 support to the virtio blk and scsi drivers in seabios. With this series applied seabios happily boots in virtio 1.0 mode from both transitional and modern devices. This series also moves all virtio code to 32bit. Tested with Fedora 22 guest, booting from virtio-scsi cdrom (live iso), virtio-scsi disk and virtio-blk disk. The patches are also available in the git repository at: git://git.kraxel.org/seabios virtio v3 changes: * change vp_device allocation. * fix capability detection. * add some cleanup patches (drop MAKESEGMENT and GET_* macros). * allocate virt queues in high memory. v2 changes: * rename vp_modern_{read_write} to vp_{read,write} * switch legacy virtio code to vp_{read,write} too. * make vp_read return the values. Gerd Hoffmann (25): pci: allow to loop over capabilities virtio: run drivers in 32bit mode virtio: add struct vp_device virtio: pass struct pci_device to vp_init_simple virtio: add version 1.0 structs and #defines virtio: add version 0.9.5 struct virtio: find version 1.0 virtio capabilities virtio: create vp_cap struct for legacy bar virtio: add read/write functions and macros virtio: make features 64bit, support version 1.0 features virtio: add version 1.0 support to vp_{get,set}_status virtio: add version 1.0 support to vp_get_isr virtio: add version 1.0 support to vp_reset virtio: add version 1.0 support to vp_notify virtio: remove unused vp_del_vq virtio: add version 1.0 support to vp_find_vq virtio-scsi: fix initialization for version 1.0 virtio-blk: fix initialization for version 1.0 virtio: use version 1.0 if available (flip the big switch) virtio: also probe version 1.0 pci ids virtio: legacy cleanup virtio-blk: 32bit cleanup virtio-scsi: 32bit cleanup virtio-ring: 32bit cleanup virtio-pci: use high memory for rings Makefile | 2 +- src/block.c | 8 +- src/fw/pciinit.c | 4 +- src/hw/blockcmd.c | 5 +- src/hw/pci.c | 11 ++- src/hw/pci.h | 2 +- src/hw/pci_ids.h | 8 +- src/hw/virtio-blk.c | 112 ++++++++++++++++------- src/hw/virtio-pci.c | 228 +++++++++++++++++++++++++++++++++++++++++----- src/hw/virtio-pci.h | 252 ++++++++++++++++++++++++++++++++++++--------------- src/hw/virtio-ring.c | 65 +++++++------ src/hw/virtio-ring.h | 9 +- src/hw/virtio-scsi.c | 75 ++++++++++----- 13 files changed, 582 insertions(+), 199 deletions(-) -- 1.8.3.1

2 27

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

SeaBIOS July 2015