Since vmd can load different firmware files, if we want to try anything
that might be considered more disruptive, we could add regular-installable
packages (i.e. handled by pkg_add not fw_update) which users could
configure per-VM. (I already do this locally for testing with a version
with the messages that Reyk doesn't like added back in ;-)
In this case I think that would be overkill but worth bearing in mind for
other changes that need testing.
--
Sent from a phone, apologies for poor formatting.
On 10 December 2018 09:23:42 Mike Larkin <mlarkin@azathoth.net> wrote:
> On Mon, Dec 10, 2018 at 09:53:15AM +0100, Claudio Jeker wrote:
>> On Sun, Dec 09, 2018 at 09:32:25PM -0800, Mike Larkin wrote:
>> > On Sun, Dec 09, 2018 at 12:19:46PM +0100, Claudio Jeker wrote:
>> > > I started looking at supporting fw_cfg in vmd. Now to make this work with
>> > > SeaBIOS there are a few fixes needed. First of all the way it reads the
>> > > FW_CFG_DATA port is not supported by vmm(4) (problem whith 'rep insb' on
>> > > IO ports). Additionally this cleans up some of the patches which I think
>> > > are not needed if the setting of screen-and-debug is changed. At least I
>> > > no longer see double printing of messages.
>> > > I enabled CONFIG_BOOTORDER because this is what I would like to tweak.
>> > > The CONFIG_DEBUG_LEVEL can be taken out before commit but it helped me
>> > > debugging this. Once vmd uses fw_cfg more of the patches can be removed.
>> > >
>> > > --
>> > > :wq Claudio
>> > >
>> >
>> > While I am not opposed to vmm-firmware updates in general, we have to be
>> > careful a bit now since there are tons of people using all sorts of bizarre
>> > linux distributions, which all do different things with the bios during boot.
>> >
>> > I'd think we would need to put this out there for a long while before I'd
>> > feel comfortable turning it on for everyone.
>>
>> I think it is the right time now to make this available as a snapshot
>> package so that people can start testing this the easy way. At least this
>> way we get a few month of full exposure before 6.5. If there are problems
>> with this we can still revert back.
>>
>> > I did not read the diff yet but I will.
>>
>> I attached an updated diff of what I plan to commit.
>>
>> --
>> :wq Claudio
>>
>
> If we've tested at least the common OpenBSD scenarios and at least one
> Linux guest, I'm ok with rolling the dice on this one.
>
> Thanks.
>
> -ml
>
>> Index: Makefile
>> ===================================================================
>> RCS file: /cvs/ports/sysutils/firmware/vmm/Makefile,v
>> retrieving revision 1.15
>> diff -u -p -r1.15 Makefile
>> --- Makefile 21 Nov 2018 00:26:05 -0000 1.15
>> +++ Makefile 9 Dec 2018 11:09:05 -0000
>> @@ -13,7 +13,7 @@ CC = /usr/bin/gcc
>> FW_DRIVER= vmm
>> FW_VER= 1.11.0
>> SB_VER= 20100422
>> -REVISION= 0
>> +REVISION= 1
>> DISTNAME= seabios-${FW_VER}
>> DISTFILES= ${DISTNAME}${EXTRACT_SUFX} \
>> sgabios-20100422{23d474943dcd55d0550a3d20b3d30e9040a4f15b}.tar.gz:0
>> Index: files/config
>> ===================================================================
>> RCS file: /cvs/ports/sysutils/firmware/vmm/files/config,v
>> retrieving revision 1.6
>> diff -u -p -r1.6 config
>> --- files/config 11 Jul 2018 09:09:46 -0000 1.6
>> +++ files/config 9 Dec 2018 12:46:20 -0000
>> @@ -15,7 +15,7 @@ CONFIG_QEMU_HARDWARE=y
>> # CONFIG_THREADS is not set
>> # CONFIG_RELOCATE_INIT is not set
>> # CONFIG_BOOTMENU is not set
>> -# CONFIG_BOOTORDER is not set
>> +CONFIG_BOOTORDER=y
>> # CONFIG_ENTRY_EXTRASTACK is not set
>> CONFIG_MALLOC_UPPERMEMORY=y
>> CONFIG_ROM_SIZE=0
>> Index: patches/patch-src_fw_paravirt_c
>> ===================================================================
>> RCS file: /cvs/ports/sysutils/firmware/vmm/patches/patch-src_fw_paravirt_c,v
>> retrieving revision 1.2
>> diff -u -p -r1.2 patch-src_fw_paravirt_c
>> --- patches/patch-src_fw_paravirt_c 26 Apr 2018 12:23:32 -0000 1.2
>> +++ patches/patch-src_fw_paravirt_c 9 Dec 2018 10:55:37 -0000
>> @@ -1,18 +1,27 @@
>> $OpenBSD: patch-src_fw_paravirt_c,v 1.2 2018/04/26 12:23:32 sthen Exp $
>>
>> +Don't use 'rep insb' to read IO ports, vmm does not support that yet.
>> Allow detection of >4GB RAM. Normally seabios only allows this with the
>> qemu config device which VMM doesn't have.
>>
>> Index: src/fw/paravirt.c
>> --- src/fw/paravirt.c.orig
>> +++ src/fw/paravirt.c
>> -@@ -500,6 +500,18 @@ qemu_cfg_e820(void)
>> - dprintf(1, "RamSizeOver4G: 0x%016llx [cmos]\n", RamSizeOver4G);
>> +@@ -257,7 +257,9 @@ qemu_cfg_read(void *buf, int len)
>> + if (qemu_cfg_dma_enabled()) {
>> + qemu_cfg_dma_transfer(buf, len, QEMU_CFG_DMA_CTL_READ);
>> + } else {
>> +- insb(PORT_QEMU_CFG_DATA, buf, len);
>> ++ u8 *d = buf;
>> ++ while (len-- > 0)
>> ++ *d++ = inb(PORT_QEMU_CFG_DATA);
>> + }
>> }
>>
>> -+void
>> -+vmm_check_high_mem(void)
>> -+{
>> +@@ -491,6 +493,18 @@ qemu_cfg_e820(void)
>> + e820_add(0xfffbc000, 4*4096, E820_RESERVED);
>> + }
>> +
>> + // Check for memory over 4Gig in cmos
>> + u64 high = ((rtc_read(CMOS_MEM_HIGHMEM_LOW) << 16)
>> + | ((u32)rtc_read(CMOS_MEM_HIGHMEM_MID) << 24)
>> @@ -22,6 +31,9 @@ Index: src/fw/paravirt.c
>> + dprintf(1, "RamSizeOver4G: 0x%016llx [cmos]\n", RamSizeOver4G);
>> +}
>> +
>> - // Populate romfile entries for legacy fw_cfg ports (that predate the
>> - // "file" interface).
>> - static void
>> ++void
>> ++vmm_check_high_mem(void)
>> ++{
>> + // Check for memory over 4Gig in cmos
>> + u64 high = ((rtc_read(CMOS_MEM_HIGHMEM_LOW) << 16)
>> + | ((u32)rtc_read(CMOS_MEM_HIGHMEM_MID) << 24)
>> Index: patches/patch-src_optionroms_c
>> ===================================================================
>> RCS file: /cvs/ports/sysutils/firmware/vmm/patches/patch-src_optionroms_c,v
>> retrieving revision 1.2
>> diff -u -p -r1.2 patch-src_optionroms_c
>> --- patches/patch-src_optionroms_c 19 Jul 2017 19:33:51 -0000 1.2
>> +++ patches/patch-src_optionroms_c 9 Dec 2018 10:45:01 -0000
>> @@ -2,9 +2,11 @@ $OpenBSD: patch-src_optionroms_c,v 1.2 2
>>
>> Needed for SGABIOS option ROM for VMM. Normally these are setup based on
>> the qemu fw_cfg interface (or coreboot CBFS on hardware).
>> +Turn of screen-and-debug to prevent double printing of chars.
>>
>> ---- src/optionroms.c.orig Fri Mar 31 09:34:40 2017
>> -+++ src/optionroms.c Fri Mar 31 09:35:35 2017
>> +Index: src/optionroms.c
>> +--- src/optionroms.c.orig
>> ++++ src/optionroms.c
>> @@ -361,7 +361,7 @@ optionrom_setup(void)
>>
>> // All option roms found and deployed - now build BEV/BCV vectors.
>> @@ -14,3 +16,12 @@ the qemu fw_cfg interface (or coreboot C
>> while (pos < rom_get_last()) {
>> struct rom_header *rom = (void*)pos;
>> if (! is_valid_rom(rom)) {
>> +@@ -413,7 +413,7 @@ vgarom_setup(void)
>> + EnforceChecksum = romfile_loadint("etc/optionroms-checksum", 1);
>> + S3ResumeVga = romfile_loadint("etc/s3-resume-vga-init", CONFIG_QEMU);
>> + RunPCIroms = romfile_loadint("etc/pci-optionrom-exec", 2);
>> +- ScreenAndDebug = romfile_loadint("etc/screen-and-debug", 1);
>> ++ ScreenAndDebug = romfile_loadint("etc/screen-and-debug", 0);
>> +
>> + // Clear option rom memory
>> + memset((void*)BUILD_ROM_START, 0, rom_get_max() - BUILD_ROM_START);
>>
No comments:
Post a Comment