- OpenBIOS - mail.coreboot.org

[PATCH] config/scripts/switch-arch: fix build on aarch64 and riscv64 hosts
by Mark Cave-Ayland July 4, 2026

July 4, 2026

Ensure that the host compiler builds with -DNATIVE_BITWIDTH_SMALLER_THAN_HOST_BITWIDTH on aarch64 and riscv64 hosts for 32-bit target architectures, otherwise host pointers can end up being truncated to 32-bits causing a segfault running forthstrap. Resolves: https://github.com/openbios/openbios/issues/18 Resolves: https://github.com/openbios/openbios/issues/25 Closes: https://github.com/openbios/openbios/pull/12 Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> --- config/scripts/switch-arch | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/config/scripts/switch-arch b/config/scripts/switch-arch index 013116b..2c20d27 100755 --- a/config/scripts/switch-arch +++ b/config/scripts/switch-arch @@ -38,7 +38,8 @@ longbits() if test "$cpu" = "sparc64" -o "$cpu" = "ia64" \ -o "$cpu" = "amd64" -o "$cpu" = "x86_64" \ -o "$cpu" = "powerpc64" -o "$cpu" = "ppc64" \ - -o "$cpu" = "ppc64le" -o "$cpu" = "alpha" ; then + -o "$cpu" = "ppc64le" -o "$cpu" = "alpha" \ + -o "$cpu" = "aarch64" -o "$cpu" = "riscv64" ; then echo 64 else echo 32 -- 2.47.3

1 1

[PATCH] .github/workflows: update actions to use latest versions
by Mark Cave-Ayland July 4, 2026

July 4, 2026

The logs for the CI builds are indicating that older actions based upon Node 20 are deprecated. Update the following versions: actions/checkout@v4 -> actions/checkout@v7 actions/download-artifact@v4 -> actions/download-artifact@v8 actions/upload-artifact@v4 -> actions/upload-artifact@v7 docker/build-push-action@v5 -> docker/build-push-action@v7 docker/login-action@v3 -> docker/login-action@v4 docker/metadata-action@v5 -> docker/metadata-action@v6 Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk> --- .github/workflows/build-openbios-builder.yml | 8 ++++---- .github/workflows/main.yml | 6 +++--- .github/workflows/release.yml | 6 +++--- 3 files changed, 10 insertions(+), 10 deletions(-) diff --git a/.github/workflows/build-openbios-builder.yml b/.github/workflows/build-openbios-builder.yml index 9715e50..93b7531 100644 --- a/.github/workflows/build-openbios-builder.yml +++ b/.github/workflows/build-openbios-builder.yml @@ -17,10 +17,10 @@ jobs: steps: - name: Checkout repository - uses: actions/checkout@v4 + uses: actions/checkout@v7 - name: Log in to the Container registry - uses: docker/login-action@v3 + uses: docker/login-action@v4 with: registry: ${{ env.REGISTRY }} username: ${{ github.actor }} @@ -28,12 +28,12 @@ jobs: - name: Extract metadata (tags, labels) for Docker id: meta - uses: docker/metadata-action@v5 + uses: docker/metadata-action@v6 with: images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }} - name: Build and push Docker image - uses: docker/build-push-action@v5 + uses: docker/build-push-action@v7 with: context: . file: docker/Dockerfile.builder diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml index a903697..9eadfc2 100644 --- a/.github/workflows/main.yml +++ b/.github/workflows/main.yml @@ -16,7 +16,7 @@ jobs: image: ghcr.io/openbios/openbios-builder:master steps: - name: Checkout OpenBIOS - uses: actions/checkout@v4 + uses: actions/checkout@v7 - name: Backup Makefile.target run: cp Makefile.target Makefile.target.orig @@ -44,7 +44,7 @@ jobs: run: "mkdir -p release && mv obj-* release" - name: Store artifacts - uses: actions/upload-artifact@v4 + uses: actions/upload-artifact@v7 with: name: openbios-multiarch-latest path: | @@ -84,7 +84,7 @@ jobs: release/obj-sparc64/QEMU,VGA.bin - name: Prepare pre-release from artifacts - uses: actions/download-artifact@v4 + uses: actions/download-artifact@v8 with: name: openbios-multiarch-latest path: archive diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml index a8b8895..bdfe932 100644 --- a/.github/workflows/release.yml +++ b/.github/workflows/release.yml @@ -16,7 +16,7 @@ jobs: image: ghcr.io/openbios/openbios-builder:master steps: - name: Checkout OpenBIOS - uses: actions/checkout@v4 + uses: actions/checkout@v7 - name: Backup Makefile.target run: cp Makefile.target Makefile.target.orig @@ -44,7 +44,7 @@ jobs: run: "mkdir -p release && mv obj-* release" - name: Store artifacts - uses: actions/upload-artifact@v4 + uses: actions/upload-artifact@v7 with: name: "openbios-multiarch-${{ github.ref_name }}" path: | @@ -84,7 +84,7 @@ jobs: release/obj-sparc64/QEMU,VGA.bin - name: Prepare release from artifacts - uses: actions/download-artifact@v4 + uses: actions/download-artifact@v8 with: name: "openbios-multiarch-${{ github.ref_name }}" path: archive -- 2.47.3

1 1

stack overflow/wraparound in 64 bit powerpc openfirmware
by Adrian Chadd June 23, 2026

June 23, 2026

hi! I spent some time debugging why FreeBSD isn't booting on qemu+mac99+openbios openfirmware. I filed an issue (https://github.com/openbios/openbios/issues/27) and a pull request (https://github.com/openbios/openbios/pull/28) . The TL;DR - SAVE_SPACE isn't enough to store all the registers - r29, r30, r31 wrap around to address 0x0 which corrupts that memory. It doesn't crash in early boot because FreeBSD hasn't yet remapped the page tables to memory mapped at 0x0 but once it does, the next openfirmware call traps. The fix is pretty simple: diff --git a/arch/ppc/qemu/start.S b/arch/ppc/qemu/start.S index c679230..4f4678f 100644 --- a/arch/ppc/qemu/start.S +++ b/arch/ppc/qemu/start.S @@ -503,7 +503,7 @@ _GLOBAL(saved_stack): #define SAVE_SPACE 320 #else #define STKOFF 8 -#define SAVE_SPACE 144 +#define SAVE_SPACE 156 #endif GLOBL(of_client_callback): Thanks! -adrian

2 2

[PATCH v2] powernv: boot OpenBSD on POWER9
by Kirill A. Korinsky June 17, 2026

June 17, 2026

OpenBSD/powernv programs PTCR for LPID 0 with a partition table size exponent one smaller than QEMU's existing ISA v3.0 interpretation. Try QEMU's existing PATS interpretation first, then fall back to the OpenBSD LPID 0 form; nonzero LPIDs keep the old behaviour. The PSI model now exposes POWER9 IRQ level and pending status registers, and keeps both updated while delivering through the existing XIVE LSI source. This lets guests that select the POWER9 PSI LSI IRQ method continue to receive LPC interrupts. The blast radius is probably minimal: the partition table fallback is limited to bare metal LPID 0 after the original lookup fails, while the PSI change only touches POWER9 PSI state and reuses the existing delivery path. Signed-off-by: Kirill A. Korinsky <kirill(a)korins.ky> --- hw/ppc/pnv_psi.c | 17 ++++----- target/ppc/mmu-book3s-v3.c | 33 +++++++++++++--- tests/functional/ppc64/meson.build | 2 + tests/functional/ppc64/test_openbsd.py | 52 ++++++++++++++++++++++++++ 4 files changed, 90 insertions(+), 14 deletions(-) create mode 100755 tests/functional/ppc64/test_openbsd.py diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c index e8701c6100..941540df2d 100644 --- a/hw/ppc/pnv_psi.c +++ b/hw/ppc/pnv_psi.c @@ -688,6 +688,8 @@ static uint64_t pnv_psi_p9_mmio_read(void *opaque, hwaddr addr, unsigned size) case PSIHB9_ESB_CI_BASE: case PSIHB9_ESB_NOTIF_ADDR: case PSIHB9_IVT_OFFSET: + case PSIHB9_IRQ_LEVEL: + case PSIHB9_IRQ_STAT: val = psi->regs[reg]; break; default: @@ -817,18 +819,15 @@ static const MemoryRegionOps pnv_psi_p9_xscom_ops = { static void pnv_psi_power9_set_irq(void *opaque, int irq, int state) { PnvPsi *psi = opaque; - uint64_t irq_method = psi->regs[PSIHB_REG(PSIHB9_INTERRUPT_CONTROL)]; + uint64_t irq_bit = PPC_BIT(irq); - if (irq_method & PSIHB9_IRQ_METHOD) { - qemu_log_mask(LOG_GUEST_ERROR, "PSI: LSI IRQ method no supported\n"); - return; - } - - /* Update LSI levels */ + /* Update LSI levels and pending status */ if (state) { - psi->regs[PSIHB_REG(PSIHB9_IRQ_LEVEL)] |= PPC_BIT(irq); + psi->regs[PSIHB_REG(PSIHB9_IRQ_LEVEL)] |= irq_bit; + psi->regs[PSIHB_REG(PSIHB9_IRQ_STAT)] |= irq_bit; } else { - psi->regs[PSIHB_REG(PSIHB9_IRQ_LEVEL)] &= ~PPC_BIT(irq); + psi->regs[PSIHB_REG(PSIHB9_IRQ_LEVEL)] &= ~irq_bit; + psi->regs[PSIHB_REG(PSIHB9_IRQ_STAT)] &= ~irq_bit; } qemu_set_irq(psi->qirqs[irq], state); diff --git a/target/ppc/mmu-book3s-v3.c b/target/ppc/mmu-book3s-v3.c index 3865556310..4babe4c536 100644 --- a/target/ppc/mmu-book3s-v3.c +++ b/target/ppc/mmu-book3s-v3.c @@ -23,19 +23,21 @@ #include "mmu-hash64.h" #include "mmu-book3s-v3.h" -bool ppc64_v3_get_pate(PowerPCCPU *cpu, target_ulong lpid, ppc_v3_pate_t *entry) +static bool ppc64_v3_get_pate_from_size(PowerPCCPU *cpu, target_ulong lpid, + ppc_v3_pate_t *entry, + uint64_t table_size) { uint64_t patb = cpu->env.spr[SPR_PTCR] & PTCR_PATB; - uint64_t pats = cpu->env.spr[SPR_PTCR] & PTCR_PATS; + uint64_t entries; /* Check if partition table is properly aligned */ - if (patb & MAKE_64BIT_MASK(0, pats + 12)) { + if (patb & (table_size - 1)) { return false; } /* Calculate number of entries */ - pats = 1ull << (pats + 12 - 4); - if (pats <= lpid) { + entries = table_size / sizeof(*entry); + if (entries <= lpid) { return false; } @@ -45,3 +47,24 @@ bool ppc64_v3_get_pate(PowerPCCPU *cpu, target_ulong lpid, ppc_v3_pate_t *entry) entry->dw1 = ldq_phys(CPU(cpu)->as, patb + 8); return true; } + +bool ppc64_v3_get_pate(PowerPCCPU *cpu, target_ulong lpid, ppc_v3_pate_t *entry) +{ + uint64_t pats = cpu->env.spr[SPR_PTCR] & PTCR_PATS; + + /* + * Keep the existing ISA v3.0 PATS interpretation first. OpenBSD/powernv + * uses the PATSIZE value it writes to PTCR as one exponent smaller, and it + * only needs that interpretation for the bare metal LPID 0 table. + */ + if (ppc64_v3_get_pate_from_size(cpu, lpid, entry, 1ull << (pats + 12))) { + return true; + } + + if (lpid == 0) { + return ppc64_v3_get_pate_from_size(cpu, lpid, entry, + 1ull << (pats + 11)); + } + + return false; +} diff --git a/tests/functional/ppc64/meson.build b/tests/functional/ppc64/meson.build index f0f8ab8f61..cb3c745624 100644 --- a/tests/functional/ppc64/meson.build +++ b/tests/functional/ppc64/meson.build @@ -4,6 +4,7 @@ test_ppc64_timeouts = { 'fadump' : 480, 'hv' : 1000, 'mac99' : 120, + 'openbsd' : 240, 'powernv' : 480, 'pseries' : 480, 'replay' : 210, @@ -20,6 +21,7 @@ tests_ppc64_system_thorough = [ 'fadump', 'hv', 'mac99', + 'openbsd', 'powernv', 'pseries', 'replay', diff --git a/tests/functional/ppc64/test_openbsd.py b/tests/functional/ppc64/test_openbsd.py new file mode 100755 index 0000000000..bdbef6bf82 --- /dev/null +++ b/tests/functional/ppc64/test_openbsd.py @@ -0,0 +1,52 @@ +#!/usr/bin/env python3 +# +# Test that OpenBSD boots on a ppc powernv machine and reaches the installer. +# +# SPDX-License-Identifier: GPL-2.0-or-later + +from qemu_test import QemuSystemTest, Asset +from qemu_test import wait_for_console_pattern + + +class OpenBSDPowerNV(QemuSystemTest): + + ASSET_MINIROOT = Asset( + 'https://kirill.korins.ky/pub/qemu-powerpc64-openbsd/miniroot79.img', + '7829e42b75d81cafd732038b9d63228b79c1f5828d8375872a4bb655e1d6b13c') + + ASSET_BOOTKERNEL = Asset( + 'https://kirill.korins.ky/pub/qemu-powerpc64-openbsd/pnor.BOOTKERNEL', + '397ce43ce61910e1a2c4f13d301f957e61513a9ec5371bc3e87d3095411fae7b') + + def test_powernv9_openbsd_installer(self): + self.set_machine('powernv9') + self.require_accelerator('tcg') + + miniroot_path = self.ASSET_MINIROOT.fetch() + bootkernel_path = self.ASSET_BOOTKERNEL.fetch() + + self.vm.set_console() + self.vm.add_args('-cpu', 'power9', + '-accel', 'tcg,thread=single', + '-smp', '1,cores=1,threads=1', + '-m', '2g', + '-kernel', bootkernel_path, + '-device', + 'ich9-ahci,id=sata0,bus=pcie.0,addr=0x0', + '-drive', + f'file={miniroot_path},format=raw,if=none,' + 'id=bootdisk,snapshot=on', + '-device', + 'ide-hd,bus=sata0.0,unit=0,drive=bootdisk,' + 'bootindex=1') + self.vm.launch() + + wait_for_console_pattern(self, 'OpenBSD 7.9 (RAMDISK)', 'panic:') + wait_for_console_pattern( + self, + '(I)nstall, (U)pgrade, (A)utoinstall or (S)hell?', + 'panic:') + + +if __name__ == '__main__': + QemuSystemTest.main() -- 2.54.0

2 1

[PATCH] Keep SPARC64 firmware memory below 4G
by Kirill A. Korinsky June 13, 2026

June 13, 2026

OpenBIOS uses the top of guest RAM for its private mappings and for reverse physical claims. On sun4u guests with more than 4G, this places firmware owned memory above the range that the existing SPARC64 setup handles correctly, which trips OpenBIOS before the guest can boot. Keep the full RAM size visible to the guest, but cap the firmware allocation search limit at 4G on SPARC64 and use the same cap for the early private mapping placement. --- arch/sparc64/entry.S | 8 +++++++- libopenbios/ofmem_common.c | 18 +++++++++++++++--- 2 files changed, 22 insertions(+), 4 deletions(-) diff --git a/arch/sparc64/entry.S b/arch/sparc64/entry.S index 224a53b..7b3f87a 100644 --- a/arch/sparc64/entry.S +++ b/arch/sparc64/entry.S @@ -133,13 +133,19 @@ entry: or %g3, %g4, %g1 ! %g1 contains end of memory + mov %g1, %g4 + srlx %g1, 32, %g5 + brz %g5, 1f + nop + setx 0x100000000, %g7, %g4 +1: setx _end, %g7, %g3 set 0x7ffff, %g2 add %g3, %g2, %g3 andn %g3, %g2, %g3 setx _data, %g7, %g2 sub %g3, %g2, %g2 - sub %g1, %g2, %g2 ! %g2 = start of private memory + sub %g4, %g2, %g2 ! %g2 = start of private memory mov %g2, %l0 ! setup .data & .bss diff --git a/libopenbios/ofmem_common.c b/libopenbios/ofmem_common.c index fc19d27..ccff11d 100644 --- a/libopenbios/ofmem_common.c +++ b/libopenbios/ofmem_common.c @@ -44,6 +44,18 @@ static ucell get_ram_size( void ) return ofmem->ramsize; } +static phys_addr_t get_claim_phys_top( void ) +{ + phys_addr_t top = get_ram_size(); + +#ifdef CONFIG_SPARC64 + if (top > (phys_addr_t)0x100000000ULL) + top = (phys_addr_t)0x100000000ULL; +#endif + + return top; +} + /************************************************************************/ /* debug */ /************************************************************************/ @@ -521,7 +533,7 @@ phys_addr_t ofmem_claim_phys( phys_addr_t phys, ucell size, ucell align ) " align=" FMT_ucellx "\n", phys, size, align); - return ofmem_claim_phys_( phys, size, align, 0, get_ram_size(), 1 ); + return ofmem_claim_phys_( phys, size, align, 0, get_claim_phys_top(), 1 ); } static ucell ofmem_claim_virt_( ucell virt, ucell size, ucell align, @@ -600,7 +612,7 @@ phys_addr_t ofmem_retain( phys_addr_t phys, ucell size, ucell align ) " align=" FMT_ucellx "\n", phys, size, align); - retain_phys = ofmem_claim_phys_( phys, size, align, 0, get_ram_size(), 1 /* reverse */ ); + retain_phys = ofmem_claim_phys_( phys, size, align, 0, get_claim_phys_top(), 1 /* reverse */ ); /* Add to the retain_phys_range list */ retained->retain_phys_range[retained->numentries].next = NULL; @@ -634,7 +646,7 @@ ucell ofmem_claim( ucell addr, ucell size, ucell align ) } else { if( align < PAGE_SIZE ) align = PAGE_SIZE; - phys = ofmem_claim_phys_( -1, size, align, 0, get_ram_size(), 1 /* reverse */ ); + phys = ofmem_claim_phys_( -1, size, align, 0, get_claim_phys_top(), 1 /* reverse */ ); virt = ofmem_claim_virt_( phys, size, 0, 0, 0, 0 ); if( phys == -1 || virt == -1 ) { OFMEM_TRACE("ofmem_claim failed\n"); -- 2.54.0

1 0

[PATCH] Make CIF boot reset the machine
by Kirill A. Korinsky June 13, 2026

June 13, 2026

Open Firmware clients call the client interface boot service as a control transfer; it must not return after a normal reboot request. The current implementation only prints BOOT and returns, which makes OpenBSD/sparc64 treat the call as OF_boot() failure and enter DDB after syncing disks. --- forth/system/ciface.fs | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/forth/system/ciface.fs b/forth/system/ciface.fs index 146561a..1f991a5 100644 --- a/forth/system/ciface.fs +++ b/forth/system/ciface.fs @@ -263,7 +263,8 @@ external \ ------------------------------------------------------------- : boot ( bootspec -- ) - ." BOOT" + drop + reset-all ; : enter ( -- ) -- 2.54.0

1 0

[PATCH] Guard client catch result storage
by Kirill A. Korinsky June 13, 2026

June 13, 2026

OpenBIOS stores the catch result in args[nargs] before copying declared return values back to the client. That slot exists only when the client advertises at least one return value; calls with nret == 0 provide no space for it. Avoid writing past the client argument frame when no return slot exists. This prevents stack and argument corruption during OpenBSD/sparc64 boot, where set-symbol-lookup is issued with zero return values. --- libopenbios/client.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/libopenbios/client.c b/libopenbios/client.c index 8b3d582..958a57c 100644 --- a/libopenbios/client.c +++ b/libopenbios/client.c @@ -272,8 +272,9 @@ handle_calls(prom_args_t *pb) return 0; } - /* Store catch result */ - pb->args[pb->nargs] = val; + /* Store catch result if the client provided room for it */ + if (pb->nret) + pb->args[pb->nargs] = val; j = dstackcnt; for (i = 1; i < pb->nret; i++, j--) { -- 2.54.0

1 0

G5 broken in the master?
by Jd Lyons March 16, 2026

March 16, 2026

Did G5 hangs come to the Openbios master, all my later builds hang the the G5 CPU, what went wrong?

3 7

Re: mac99 SMP
by Jd Lyons Jan. 15, 2026

Jan. 15, 2026

What did you do Howard? Checking my work to get OS 9 not the freeze on 2nd CPU access I think I gave it the wrong interrupt-parent I give it Mac-io>interrupt-controller and it should have been uni-north as that controls interrupts first. I think making this change to Openbios for the CPU properties will fix OS 9! > On Dec 19, 2025, at 9:05 AM, Howard Spoelstra <hsp.cat7(a)gmail.com> wrote: > > Hello all, > > Long time no write. Here are just two teasers ;-) > > >

1 0

Re: PCI Passthrough of GPUs
by Jd Lyons Nov. 24, 2025

Nov. 24, 2025

Removing the NVDAResman.kext will get us to the desktop, even tho the bad range is reported for the BARs above 0x9xxxxxxx is’t happy to use them Nov 14 12:13:24 kextd[28]: 0 cached, 218 uncached personalities to catalog ndrv base = 4276e000 ndrv base = 41a6a000 Nov 14 12:13:25 localhost DirectoryService[48]: Launched version 2.1 (v353.6) Nov 14 12:13:25 localhost mDNSResponder-108.6 (Jul 19 2007 11: 33:32)[38]: starting UniNEnet: Ethernet address 52:54:00:12:34:56 Nov 14 12:13:26 localhost /System/Library/CoreServices/loginwindow.app/Contents/MacOS/loginwindow: Login Window Application Started Attached protocol 30 to en0 - 0 UniNEnet::monitorLinkStatus - Link is up at 100 Mbps - Full Duplex Attached protocol 2 to en0 - 0 NVDA,Display-A: vram [90020000:08000000] NVDA,Display-B: vram [92000000:08000000] initialize_screen: b=4391B000, w=00000F00, h=00000870, r=00004000 initialize_screen: b=4391B000, w=00000F00, h=00000870, r=00004000 kPEEnableScreen -1 in.c: warning can't plumb proto if=en0 type 6 error=17 > On Nov 14, 2025, at 10:14 AM, Jd Lyons <lyons_dj(a)yahoo.com> wrote: > > > >> On Nov 14, 2025, at 10:07 AM, BALATON Zoltan <balaton(a)eik.bme.hu> wrote: >> >> On Fri, 14 Nov 2025, Jd Lyons via OpenBIOS wrote: >>> I finally got this working with a GF7800GT and it boots linux on G4/G5 on mac99 OpenGL seems to work. >>> >>> Having an issue with mac99 only having a 256MB PCI hole window and we need more than that for 256+ MB VRAM cards. I patched it in openbios, but the uninorth.c in qemu says 256MB is the real PCI memory hole on mac99?? >> >> I was told here: >> https://www.amigans.net/modules/newbb/viewtopic.php?post_id=141184#forumpos… >> that this may be normal for old machines. Maybe you need to disable resizable BARs in your BIOS/UEFI setup of the host? > > > I did disable resizable bars some time back. >> >>> I patched it to 1GB but OS X still has issues with it and kernel panics when the graphics come up to the desktop. >> >> This may not work as that could overlap other I/O areas. I think the driver should be able to handle the 256MB BAR limitation as that seems to be normal for these cards. >> > > These cards have 64bit BARs it moves it outside the 265MB range with 265MB of VRAM alone. No? > > > >> Regards, >> BALATON Zoltan

3 5