Recent Posts

21

Operating Systems and Porting / Re: Debian 12 status?

« Last post by pocock on July 03, 2024, 11:16:52 am »

I took the 6.7.12-1 backport for bookworm and applied the 4k page size patch.

It has the same problem with amdgpu that I saw using the 6.1 kernel with 4k patch.  Errors about "Not enough memory for command submission!" and eventually Firefox and the whole desktop freeze.

I pushed each of my branches for anybody else who wants to try it.  Maybe it is fine with other GPUs

This is the 4k patch against the 6.7.12-1 kernel on the branch for unstable and testing

The same patch backported for the 6.7.12-1 kernel on bookworm-backports

The 6.1 standard kernel branch for bookworm with the 4k page size patch

There are various other web sites discussing the error message.  Some of them suggest that the amdgpu firmware needs to be updated.

One of them points to a new version of the amdgpu firmware that appeared in the upstream kernel tree in June 2024

At present, the Debian firmware package still has older amdgpu firmware from 2023 - you can see the current versions here, even unstable hasn't been updated since June 2023.

Here is the packaging repository on Debian Git (Salsa), we can't clearly see the actual versions of the amdgpu firmware files

22

Operating Systems and Porting / Re: [NEWS] Fedora 40 is officially released!

« Last post by ClassicHasClass on July 02, 2024, 05:49:54 pm »

That gives me some point of comparison, though it may be difficult to figure out what specifically the ARM dynrec is benefiting from. But thanks for looking at it.

23

Operating Systems and Porting / Re: Debian 12 status?

« Last post by pocock on July 02, 2024, 05:02:25 pm »

Using the Debian 12 (bookworm) kernel 6.1 package with 4k page size I started getting errors in the log and eventually Firefox would freeze and the whole GUI (GNOME desktop) would freeze.  The system is still accessible over SSH.

After the first crash, I started radeontop and noticed up to 80% GPU VRAM utilisation with Firefox running.

I rebooted into the kernel from bullseye (5.10.46-4.1 built from my 4k page size branch), running with the Debian 12 filesystem and it is running fine, no crashes.  I have had that kernel running for months on end on this platform using the bullseye filesystem.

I'm going to build the 6.7.12 kernel backport for bookworm with the 4k page size and try that as well.

journalctl captures a lot of errors like this, sometimes they appear for hours before it eventually crashes.  I was able to repeat the crash a couple of times.

I see it as a good thing that the platform is still responding over SSH even when the GUI has crashed.

Code: [Select]

kernel: [drm:amdgpu_cs_ioctl [amdgpu]] *ERROR* Not enough memory for command submission!
kernel: [drm:amdgpu_cs_ioctl [amdgpu]] *ERROR* Not enough memory for command submission!
kernel: [drm:amdgpu_cs_ioctl [amdgpu]] *ERROR* Not enough memory for command submission!
...

The crash is captured too:

Code: [Select]

kernel: ------------[ cut here ]------------
 kernel: WARNING: CPU: 5 PID: 6577 at drivers/gpu/drm/ttm/ttm_bo.c:357 ttm_bo_release+0x538/0x5b0 [ttm]
 kernel: Modules linked in: xt_conntrack nft_chain_nat xt_MASQUERADE nf_nat nf_conntrack_netlink nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 xfrm_user xfrm_algo xt_addrtype nft_compat nf_tables nfnetlink br_netfilter bridge snd_seq_dummy snd_hrtimer snd_seq rfkill qrtr 8021q garp stp mrp llc overlay sunrpc binfmt_misc ext4 crc16 mbcache jbd2 uvcvideo videobuf2_vmalloc videobuf2_memops snd_usb_audio snd_hda_codec_hdmi videobuf2_v4l2 videobuf2_common snd_usbmidi_lib snd_hda_intel snd_rawmidi snd_intel_dspcfg videodev snd_seq_device evdev joydev snd_hda_codec mc snd_hda_core snd_hwdep snd_pcm sg snd_timer snd ofpart soundcore ipmi_powernv powernv_flash ctr ipmi_devintf at24 vmx_crypto mtd regmap_i2c ipmi_msghandler gf128mul opal_prd parport_pc lp parport fuse configfs loop ip_tables x_tables autofs4 btrfs blake2b_generic zstd_compress xts ecb uas usb_storage dm_crypt dm_mod raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor raid6_pq libcrc32c crc32c_generic
 kernel:  raid1 raid0 multipath linear md_mod hid_generic usbhid hid sd_mod amdgpu gpu_sched drm_buddy i2c_algo_bit drm_display_helper drm_ttm_helper ttm drm_kms_helper syscopyarea sysfillrect sysimgblt xhci_pci fb_sys_fops xhci_hcd nvme nvme_core t10_pi tg3 drm mpt3sas crc64_rocksoft_generic usbcore crc64_rocksoft crc_t10dif crct10dif_generic crc64 crct10dif_common drm_panel_orientation_quirks raid_class libphy usb_common scsi_transport_sas
 kernel: CPU: 5 PID: 6577 Comm: Renderer Not tainted 6.1.0-21-powerpc64le-4k #1  Debian 6.1.90-1.1
 kernel: Hardware name: T2P9D01 REV 1.00 POWER9 0x4e1203 opal:skiboot-9858186 PowerNV
 kernel: NIP:  c00800000f4d2120 LR: c008000012fe7270 CTR: c00800000f4d2198
 kernel: REGS: c00020001bc571d0 TRAP: 0700   Not tainted  (6.1.0-21-powerpc64le-4k Debian 6.1.90-1.1)
 kernel: MSR:  9000000000029033 <SF,HV,EE,ME,IR,DR,RI,LE>  CR: 84824244  XER: 20040036
 kernel: CFAR: c00800000f4d1c34 IRQMASK: 0
         GPR00: c008000012fe7270 c00020001bc57470 c00800000f508800 c0000003b100c5c0
         GPR04: 0000000000000000 0000000ffcac0000 0000000000041a8b 0000000000000000
         GPR08: 0000000000041a8a c00020001af87738 0000000000000000 c008000013564c90
         GPR12: c00800000f4d2198 c000000ffffdbf00 c00020001b054b80 c0002000ef821798
         GPR16: 00000000002c6800 0000000000000001 0000000000000000 0000000000000000
         GPR20: 00000000002c6880 c000200015080000 0000000000000071 00000000002c6800
         GPR24: 0000000000000003 c00020001af87010 0000000000000000 000000003ee00000
         GPR28: c0000003b100c458 c000200015080000 c000200015085508 c0000003b100c5c0
 kernel: NIP [c00800000f4d2120] ttm_bo_release+0x538/0x5b0 [ttm]
 kernel: LR [c008000012fe7270] amdgpu_bo_unref+0x38/0x60 [amdgpu]
 kernel: Call Trace:
 kernel: [c00020001bc57470] [c00800000f4d1f18] ttm_bo_release+0x330/0x5b0 [ttm] (unreliable)
 kernel: [c00020001bc57500] [c008000012fe7270] amdgpu_bo_unref+0x38/0x60 [amdgpu]
 kernel: [c00020001bc57530] [c0080000130105fc] amdgpu_vm_ptes_update+0xc24/0xc60 [amdgpu]
 kernel: [c00020001bc576a0] [c00800001300935c] amdgpu_vm_update_range+0x304/0x880 [amdgpu]
 kernel: [c00020001bc577c0] [c008000013009f04] amdgpu_vm_bo_update+0x2ec/0x630 [amdgpu]
 kernel: [c00020001bc578e0] [c008000012ff0bcc] amdgpu_gem_va_ioctl+0x674/0x6b0 [amdgpu]
 kernel: [c00020001bc57a20] [c008000012f07040] drm_ioctl_kernel+0x118/0x230 [drm]
 kernel: [c00020001bc57a80] [c008000012f073b0] drm_ioctl+0x258/0x560 [drm]
 kernel: [c00020001bc57bf0] [c008000012fc00b8] amdgpu_drm_ioctl+0x70/0xd0 [amdgpu]
 kernel: [c00020001bc57c40] [c0000000005493f4] sys_ioctl+0x744/0x1460
 kernel: [c00020001bc57d40] [c00000000002afd8] system_call_exception+0x138/0x260
 kernel: [c00020001bc57e10] [c00000000000c0f0] system_call_vectored_common+0xf0/0x280
 kernel: --- interrupt: 3000 at 0x7fff8eb4433c
 kernel: NIP:  00007fff8eb4433c LR: 00007fff8eb4433c CTR: 0000000000000000
 kernel: REGS: c00020001bc57e80 TRAP: 3000   Not tainted  (6.1.0-21-powerpc64le-4k Debian 6.1.90-1.1)
 kernel: MSR:  900000000280f033 <SF,HV,VEC,VSX,EE,PR,FP,ME,IR,DR,RI,LE>  CR: 44224840  XER: 00000000
 kernel: IRQMASK: 0
         GPR00: 0000000000000036 00007fff75b9bb20 00007fff8ec56f00 0000000000000053
         GPR04: 00000000c0286448 00007fff75b9bc00 0000000000280000 00000002c5a00000
         GPR08: 0000000000100000 0000000000000000 0000000000000000 0000000000000000
         GPR12: 0000000000000000 00007fff75ba68c0 00007fff75b9c028 00007fff75b9c178
         GPR16: 00007fff75b9c508 0000000000000001 0000000000020000 00007fff75b9cd18
         GPR20: 0000000000000001 0000000000020000 00007fff75b9be80 0000000000ea0000
         GPR24: 0000000000000000 0000000000200000 0000000000000004 0000000000200000
         GPR28: 0000000000000053 00000000c0286448 00007fff75b9bc00 00007ffefa5ca0a0
 kernel: NIP [00007fff8eb4433c] 0x7fff8eb4433c
 kernel: LR [00007fff8eb4433c] 0x7fff8eb4433c
 kernel: --- interrupt: 3000
 kernel: Instruction dump:
 kernel: 4bfffe30 60000000 60000000 60420000 0fe00000 7c0802a6 fb610068 fba10078
 kernel: f80100a0 60000000 60000000 60420000 <0fe00000> 4bffffe0 60000000 60420000
 kernel: ---[ end trace 0000000000000000 ]---
 kernel: [drm:amdgpu_gem_va_ioctl [amdgpu]] *ERROR* Couldn't update BO_VA (-12)
 kernel: Kernel attempted to read user page (0) - exploit attempt? (uid: 1000)
 kernel: BUG: Kernel NULL pointer dereference on read at 0x00000000
 kernel: Faulting instruction address: 0xc008000012fe8f30
 kernel: Oops: Kernel access of bad area, sig: 11 [#1]
 kernel: LE PAGE_SIZE=64K MMU=Radix SMP NR_CPUS=2048 NUMA PowerNV
 kernel: Modules linked in: xt_conntrack nft_chain_nat xt_MASQUERADE nf_nat nf_conntrack_netlink nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 xfrm_user xfrm_algo xt_addrtype nft_compat nf_tables nfnetlink br_netfilter bridge snd_seq_dummy snd_hrtimer snd_seq rfkill qrtr 8021q garp stp mrp llc overlay sunrpc binfmt_misc ext4 crc16 mbcache jbd2 uvcvideo videobuf2_vmalloc videobuf2_memops snd_usb_audio snd_hda_codec_hdmi videobuf2_v4l2 videobuf2_common snd_usbmidi_lib snd_hda_intel snd_rawmidi snd_intel_dspcfg videodev snd_seq_device evdev joydev snd_hda_codec mc snd_hda_core snd_hwdep snd_pcm sg snd_timer snd ofpart soundcore ipmi_powernv powernv_flash ctr ipmi_devintf at24 vmx_crypto mtd regmap_i2c ipmi_msghandler gf128mul opal_prd parport_pc lp parport fuse configfs loop ip_tables x_tables autofs4 btrfs blake2b_generic zstd_compress xts ecb uas usb_storage dm_crypt dm_mod raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor raid6_pq libcrc32c crc32c_generic
 kernel:  raid1 raid0 multipath linear md_mod hid_generic usbhid hid sd_mod amdgpu gpu_sched drm_buddy i2c_algo_bit drm_display_helper drm_ttm_helper ttm drm_kms_helper syscopyarea sysfillrect sysimgblt xhci_pci fb_sys_fops xhci_hcd nvme nvme_core t10_pi tg3 drm mpt3sas crc64_rocksoft_generic usbcore crc64_rocksoft crc_t10dif crct10dif_generic crc64 crct10dif_common drm_panel_orientation_quirks raid_class libphy usb_common scsi_transport_sas
 kernel: CPU: 4 PID: 6567 Comm: firefox-es:cs0 Tainted: G        W          6.1.0-21-powerpc64le-4k #1  Debian 6.1.90-1.1
 kernel: Hardware name: T2P9D01 REV 1.00 POWER9 0x4e1203 opal:skiboot-9858186 PowerNV
 kernel: NIP:  c008000012fe8f30 LR: c008000013030f54 CTR: c00000000098c350
 kernel: REGS: c000200033e2ae10 TRAP: 0300   Tainted: G        W           (6.1.0-21-powerpc64le-4k Debian 6.1.90-1.1)
 kernel: MSR:  900000000280b033 <SF,HV,VEC,VSX,EE,FP,ME,IR,DR,RI,LE>  CR: 24224244  XER: 200400dd
 kernel: CFAR: c008000013030f50 DAR: 0000000000000000 DSISR: 40000000 IRQMASK: 0
         GPR00: c008000013030f54 c000200033e2b0b0 c00800001377c600 c000200015085508
         GPR04: c0000003b100c400 0000000000000000 000000003ee00000 0000000000000080
         GPR08: 0000000000001000 0000000000000000 c0000003cce4e800 c008000013565488
         GPR12: c00000000098c350 c000000ffffdcc00 00000000002c6880 00000000002c6880
         GPR16: 0000000000080000 c0000003b100c400 0000000000001000 c000200033e2b408
         GPR20: 0000000000000000 c000200015080000 c0000003b100c400 0000000000000000
         GPR24: 000000003ee00000 c0000003cce4e800 00000000000003f1 0000000000001000
         GPR28: 000000003ee00000 c000200033e2b3e8 0000000000000080 0000000000000000
 kernel: NIP [c008000012fe8f30] amdgpu_bo_gpu_offset_no_check+0x28/0x78 [amdgpu]
 kernel: LR [c008000013030f54] amdgpu_vm_sdma_set_ptes+0x5c/0x1b0 [amdgpu]
 kernel: Call Trace:
 kernel: [c000200033e2b0b0] [c000200033e2b110] 0xc000200033e2b110 (unreliable)
 kernel: [c000200033e2b0e0] [c008000013030f54] amdgpu_vm_sdma_set_ptes+0x5c/0x1b0 [amdgpu]
 kernel: [c000200033e2b150] [c00800001303195c] amdgpu_vm_sdma_update+0x3b4/0x440 [amdgpu]
 kernel: [c000200033e2b220] [c00800001300fd40] amdgpu_vm_ptes_update+0x368/0xc60 [amdgpu]
 kernel: [c000200033e2b390] [c00800001300935c] amdgpu_vm_update_range+0x304/0x880 [amdgpu]
 kernel: [c000200033e2b4b0] [c008000013009f04] amdgpu_vm_bo_update+0x2ec/0x630 [amdgpu]
 kernel: [c000200033e2b5d0] [c008000012ff5a28] amdgpu_cs_ioctl+0x1610/0x22c0 [amdgpu]
 kernel: [c000200033e2b880] [c008000012f07040] drm_ioctl_kernel+0x118/0x230 [drm]
 kernel: [c000200033e2b8e0] [c008000012f073b0] drm_ioctl+0x258/0x560 [drm]
 kernel: [c000200033e2ba50] [c008000012fc00b8] amdgpu_drm_ioctl+0x70/0xd0 [amdgpu]
 kernel: [c000200033e2baa0] [c0000000005493f4] sys_ioctl+0x744/0x1460
 kernel: [c000200033e2bba0] [c00000000002afd8] system_call_exception+0x138/0x260
 kernel: [c000200033e2be10] [c00000000000c0f0] system_call_vectored_common+0xf0/0x280
 kernel: --- interrupt: 3000 at 0x7fff8eb4433c
 kernel: NIP:  00007fff8eb4433c LR: 00007fff8eb4433c CTR: 0000000000000000
 kernel: REGS: c000200033e2be80 TRAP: 3000   Tainted: G        W           (6.1.0-21-powerpc64le-4k Debian 6.1.90-1.1)
 kernel: MSR:  900000000280f033 <SF,HV,VEC,VSX,EE,PR,FP,ME,IR,DR,RI,LE>  CR: 48884842  XER: 00000000
 kernel: IRQMASK: 0
         GPR00: 0000000000000036 00007fff62bfe1f0 00007fff8ec56f00 0000000000000053
         GPR04: 00000000c0186444 00007fff62bfe2f0 0000000000180000 00007fff62bfe478
         GPR08: 0000000000100000 0000000000000000 0000000000000000 0000000000000000
         GPR12: 0000000000000000 00007fff62c068c0 00007fff7dc86600 0000000000000005
         GPR16: 00007fff623f0000 0000000000000001 0000000000000031 0000000000000001
         GPR20: fffffffffffffffd 0000000000000000 00007fff59870000 00007fff59860000
         GPR24: 00007fff8b4e1000 00007fff62bfe428 00007fff62bfe448 0000000000000000
         GPR28: 0000000000000053 00000000c0186444 00007fff62bfe2f0 00007fff62bfe2d0
 kernel: NIP [00007fff8eb4433c] 0x7fff8eb4433c
 kernel: LR [00007fff8eb4433c] 0x7fff8eb4433c
 kernel: --- interrupt: 3000
 kernel: Instruction dump:
 kernel: 007936f8 00000000 3c4c0079 384236f8 7c0802a6 60000000 7c0802a6 fbe1fff8
 kernel: f8010010 f821ffd1 e92301c8 e86301a8 <ebe90000> 80890010 3863aaf8 7bff83e4
 kernel: ---[ end trace 0000000000000000 ]---
 kernel:
 kernel: note: firefox-es:cs0[6567] exited with irqs disabled


24

Operating Systems and Porting / Re: [NEWS] Fedora 40 is officially released!

« Last post by Hasturtium on July 01, 2024, 08:50:56 pm »

It was not. Doing a comparison with Rosetta enabled versus ARM-native yielded this result:

Running timedemo demo1 for DOS Quake v1.07 at 320x200 in DOSBox-Staging:
Rosetta: 80.5 fps
Native: 145.0 fps

I don't know why I thought the Air pulled 200+ fps earlier, but 145 fps is easily high-end Pentium III territory regardless.

25

Operating Systems and Porting / Re: [NEWS] Fedora 40 is officially released!

« Last post by ClassicHasClass on July 01, 2024, 06:59:30 pm »

Is Rosetta checked in Get Info?

26

Operating Systems and Porting / Re: Debian 12 status?

« Last post by pocock on July 01, 2024, 10:42:28 am »

Did anybody ever ask other Debian kernel people about integrating my patch in the official kernel packages?

I've updated the 4k kernel patch to work with Debian 12

It is on this branch in a repository that is forked from the official Debian kernel packaging repository.  I've kept the change set as minimal as possible so it is very easy to diff it against the official kernel and see that it is just the page size config option changing.

To make it compile, I had to include one extra patch, the patch from Debian bug #996170.

Any feedback is welcome whether you are running the 64k or the 4k kernel

This is where I posted the full set of commands for compiling a Debian kernel with the patch

27

Operating Systems and Porting / Re: [NEWS] Fedora 40 is officially released!

« Last post by Hasturtium on July 01, 2024, 10:13:40 am »

I'm not sure, actually. I'll have to load it up on my M1 Air and check it out. The x86_64 figure could be because those don't use dynrec, but ARM should ... unless you're running the Intel Mac build on your M2, not an Apple silicon build?

You're right - I figured the Ryzen was largely running native code. To the point on Apple Silicon, the DOSBox Staging build I snagged indicated it was a Universal Binary for Apple Silicon and Intel, but I  wonder if it it's running an Intel-native binary and doing an end run around dynrec. Let me know what you find on the M1.

28

Operating Systems and Porting / Re: [NEWS] Fedora 40 is officially released!

« Last post by ClassicHasClass on June 30, 2024, 09:02:37 pm »

I'm not sure, actually. I'll have to load it up on my M1 Air and check it out. The x86_64 figure could be because those don't use dynrec, but ARM should ... unless you're running the Intel Mac build on your M2, not an Apple silicon build?

29

Operating Systems and Porting / Re: [NEWS] Fedora 40 is officially released!

« Last post by Hasturtium on June 30, 2024, 09:55:28 am »

I'm not worried, Redneck Rampage was always flaky and in hours since it hasn't so much as burbled - which means DOSBox-Staging is at least as stable as any actual DOS machine I ever ran.

Incidentally, I did a performance comparison of DOSBox-Staging between my three machines by running Quake in 320x200, and both my M2 Air and Ryzen 5700GE notch over 200 fps, while the Power9 wrangles about 35. I want to be clear that I'm not complaining: the Power9 is delivering performance that feels very much like a midrange Pentium, which means it's sufficient for anything that all but crazy people would ask of MS-DOS, but I am curious about the sheer amount of that difference.

30

Blackbird / Re: Availability of P9 hardware post-ISA 3.1 HW launch

« Last post by Hasturtium on June 29, 2024, 10:34:57 pm »

I've seen one or two motherboard/CPU combos pop up on eBay, never for more than a few days at a time. The CPUs are readily available, frequently for rates that make me seriously contemplate snagging another motherboard... until I look back and see how high the prices have risen. Oi. I love my Blackbird and use it every day, to the point that it's become my main machine unless I run into something it literally can't do, but I couldn't rationalize getting another one now.

It would not be surprising to see Blackbirds and Talos boards become more available with time - ARM workstation motherboards are becoming more abundant and performant, the platform is aging, RISC-V keeps threatening to grow into more than a cargo cult that'll re-order the embedded market. But nobody's getting mine.