From mboxrd@z Thu Jan 1 00:00:00 1970 From: bugzilla-daemon@freedesktop.org Subject: [Bug 111229] Unable to unbind GPU from amdgpu Date: Mon, 21 Oct 2019 07:22:46 +0000 Message-ID: References: Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============0987042163==" Return-path: Received: from culpepper.freedesktop.org (culpepper.freedesktop.org [IPv6:2610:10:20:722:a800:ff:fe98:4b55]) by gabe.freedesktop.org (Postfix) with ESMTP id 0B2E2899D4 for ; Mon, 21 Oct 2019 07:22:46 +0000 (UTC) In-Reply-To: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" To: dri-devel@lists.freedesktop.org List-Id: dri-devel@lists.freedesktop.org --===============0987042163== Content-Type: multipart/alternative; boundary="15716425660.B3CBC0.9711" Content-Transfer-Encoding: 7bit --15716425660.B3CBC0.9711 Date: Mon, 21 Oct 2019 07:22:45 +0000 MIME-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Bugzilla-URL: http://bugs.freedesktop.org/ Auto-Submitted: auto-generated https://bugs.freedesktop.org/show_bug.cgi?id=3D111229 --- Comment #11 from Eugene Shatsky --- Since last comment I've used this for a dozen times for switching between L= inux desktop and Windows VM, one time amdgpu crashed after resume from suspend b= ut I'm not sure if it was related to this bug and I was still able to reboot a= fter it. However I still get this warning sometimes on unbind: WARNING: CPU: 0 PID: 1109 at drivers/gpu/drm/amd/amdgpu/amdgpu_object.c:929 amdgpu_bo_unpin+0xc8/0xf0 [amdgpu] Modules linked in: vfio_pci vfio_virqfd vfio_iommu_type1 vfio fuse amdgpu amd_iommu_v2 gpu_sched ttm xt_CHECKSUM xt_MASQUERADE ipt_REJECT nf_rejec> nf_conntrack nf_defrag_ipv4 libcrc32c zsmalloc ip6t_rpfilter ipt_rpfilter ip6table_raw iptable_raw xt_pkttype nf_log_ipv6 nf_log_ipv4 nf_log_comm> CPU: 0 PID: 1109 Comm: .libvirtd-wrapp Tainted: G O 5.3.0-rc7 #1-NixOS Hardware name: To Be Filled By O.E.M. To Be Filled By O.E.M./H61M-DGS R2.0, BIOS P1.10 10/01/2013 RIP: 0010:amdgpu_bo_unpin+0xc8/0xf0 [amdgpu] Code: ff 48 83 c0 0c 48 39 d0 75 ea 48 8d 73 30 48 8d 7b 50 48 8d 54 24 08 = e8 46 1f d8 ff 85 c0 74 a1 e9 30 6c 21 00 e8 28 f9 6b f5 <0f> 0b 48 8b > RSP: 0018:ffffa4df00a4bd28 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff8c60449a4800 RCX: 0000000000000002 RDX: ffff8c60423c9b00 RSI: 0000000000000000 RDI: ffff8c60449a4800 RBP: ffff8c6008fa4058 R08: 0000000000000000 R09: ffffffffc0b3c000 R10: ffff8c60449a2800 R11: 0000000000000001 R12: ffff8c6008fa6378 R13: ffff8c6008fa6370 R14: ffff8c6008fa4058 R15: ffff8c6008d7f260 FS: 00007fac9a81f700(0000) GS:ffff8c605f400000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007ffea51ccff8 CR3: 00000004048c4003 CR4: 00000000001606f0 Call Trace: amdgpu_bo_free_kernel+0x6b/0x120 [amdgpu] amdgpu_gfx_rlc_fini+0x47/0x70 [amdgpu] gfx_v8_0_sw_fini+0xa1/0x1a0 [amdgpu] amdgpu_device_fini+0x257/0x479 [amdgpu] amdgpu_driver_unload_kms+0x4a/0x90 [amdgpu] drm_dev_unregister+0x4b/0xb0 [drm] amdgpu_pci_remove+0x25/0x50 [amdgpu] pci_device_remove+0x3b/0xc0 device_release_driver_internal+0xd8/0x1b0 unbind_store+0x94/0x120 kernfs_fop_write+0x108/0x190 vfs_write+0xa5/0x1a0 ksys_write+0x59/0xd0 do_syscall_64+0x4e/0x120 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x7faca4a7b36f Code: 1f 40 00 41 54 55 49 89 d4 53 48 89 f5 89 fb 48 83 ec 10 e8 53 fd ff = ff 4c 89 e2 41 89 c0 48 89 ee 89 df b8 01 00 00 00 0f 05 <48> 3d 00 f0 > RSP: 002b:00007fac9a81e4d0 EFLAGS: 00000293 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 0000000000000012 RCX: 00007faca4a7b36f RDX: 000000000000000c RSI: 00007fac84019a20 RDI: 0000000000000012 RBP: 00007fac84019a20 R08: 0000000000000000 R09: 000000000000002f R10: 0000000000000000 R11: 0000000000000293 R12: 000000000000000c R13: 0000000000000000 R14: 0000000000000012 R15: 00007fac9a81e568 ---[ end trace ffd153eee3d00ec4 ]--- amdgpu 0000:01:00.0: 00000000001146cc unpin not necessary It's produced by https://github.com/torvalds/linux/blob/574cc4539762561d96b456dbc0544d8898bd= 4c6e/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c#L937 , I wonder if buffer object pin count is something like reference count Also it looks like the message *ERROR* Device removal is currently not supported outside of fbcon is printed non-conditionally, without checking if DRM nodes are being used = by userspace clients. I wonder if it's possible to implement such a check and prevent the unbind if they are --=20 You are receiving this mail because: You are the assignee for the bug.= --15716425660.B3CBC0.9711 Date: Mon, 21 Oct 2019 07:22:45 +0000 MIME-Version: 1.0 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Bugzilla-URL: http://bugs.freedesktop.org/ Auto-Submitted: auto-generated

Comme= nt # 11 on bug 11122= 9 from Eugene Shatsky 
Since last comment I've used this for a dozen times for switch=
ing between Linux
desktop and Windows VM, one time amdgpu crashed after resume from suspend b=
ut
I'm not sure if it was related to this bug and I was still able to reboot a=
fter
it.
However I still get this warning sometimes on unbind:

WARNING: CPU: 0 PID: 1109 at drivers/gpu/drm/amd/amdgpu/amdgpu_object.c:929
amdgpu_bo_unpin+0xc8/0xf0 [amdgpu]
Modules linked in: vfio_pci vfio_virqfd vfio_iommu_type1 vfio fuse amdgpu
amd_iommu_v2 gpu_sched ttm xt_CHECKSUM xt_MASQUERADE ipt_REJECT nf_rejec>
 nf_conntrack nf_defrag_ipv4 libcrc32c zsmalloc ip6t_rpfilter ipt_rpfilter
ip6table_raw iptable_raw xt_pkttype nf_log_ipv6 nf_log_ipv4 nf_log_comm>
CPU: 0 PID: 1109 Comm: .libvirtd-wrapp Tainted: G           O      5.3.0-rc7
#1-NixOS
Hardware name: To Be Filled By O.E.M. To Be Filled By O.E.M./H61M-DGS R2.0,
BIOS P1.10 10/01/2013
RIP: 0010:amdgpu_bo_unpin+0xc8/0xf0 [amdgpu]
Code: ff 48 83 c0 0c 48 39 d0 75 ea 48 8d 73 30 48 8d 7b 50 48 8d 54 24 08 =
e8
46 1f d8 ff 85 c0 74 a1 e9 30 6c 21 00 e8 28 f9 6b f5 <0f> 0b 48 8b &=
gt;
RSP: 0018:ffffa4df00a4bd28 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff8c60449a4800 RCX: 0000000000000002
RDX: ffff8c60423c9b00 RSI: 0000000000000000 RDI: ffff8c60449a4800
RBP: ffff8c6008fa4058 R08: 0000000000000000 R09: ffffffffc0b3c000
R10: ffff8c60449a2800 R11: 0000000000000001 R12: ffff8c6008fa6378
R13: ffff8c6008fa6370 R14: ffff8c6008fa4058 R15: ffff8c6008d7f260
FS:  00007fac9a81f700(0000) GS:ffff8c605f400000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007ffea51ccff8 CR3: 00000004048c4003 CR4: 00000000001606f0
Call Trace:
 amdgpu_bo_free_kernel+0x6b/0x120 [amdgpu]
 amdgpu_gfx_rlc_fini+0x47/0x70 [amdgpu]
 gfx_v8_0_sw_fini+0xa1/0x1a0 [amdgpu]
 amdgpu_device_fini+0x257/0x479 [amdgpu]
 amdgpu_driver_unload_kms+0x4a/0x90 [amdgpu]
 drm_dev_unregister+0x4b/0xb0 [drm]
 amdgpu_pci_remove+0x25/0x50 [amdgpu]
 pci_device_remove+0x3b/0xc0
 device_release_driver_internal+0xd8/0x1b0
 unbind_store+0x94/0x120
 kernfs_fop_write+0x108/0x190
 vfs_write+0xa5/0x1a0
 ksys_write+0x59/0xd0
 do_syscall_64+0x4e/0x120
 entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x7faca4a7b36f
Code: 1f 40 00 41 54 55 49 89 d4 53 48 89 f5 89 fb 48 83 ec 10 e8 53 fd ff =
ff
4c 89 e2 41 89 c0 48 89 ee 89 df b8 01 00 00 00 0f 05 <48> 3d 00 f0 &=
gt;
RSP: 002b:00007fac9a81e4d0 EFLAGS: 00000293 ORIG_RAX: 0000000000000001
RAX: ffffffffffffffda RBX: 0000000000000012 RCX: 00007faca4a7b36f
RDX: 000000000000000c RSI: 00007fac84019a20 RDI: 0000000000000012
RBP: 00007fac84019a20 R08: 0000000000000000 R09: 000000000000002f
R10: 0000000000000000 R11: 0000000000000293 R12: 000000000000000c
R13: 0000000000000000 R14: 0000000000000012 R15: 00007fac9a81e568
---[ end trace ffd153eee3d00ec4 ]---
amdgpu 0000:01:00.0: 00000000001146cc unpin not necessary

It's produced by
https://gi=
thub.com/torvalds/linux/blob/574cc4539762561d96b456dbc0544d8898bd4c6e/drive=
rs/gpu/drm/amd/amdgpu/amdgpu_object.c#L937
, I wonder if buffer object pin count is something like reference count

Also it looks like the message

*ERROR* Device removal is currently not supported outside of fbcon

is printed non-conditionally, without checking if DRM nodes are being used =
by
userspace clients. I wonder if it's possible to implement such a check and
prevent the unbind if they are

You are receiving this mail because:
  • You are the assignee for the bug.
= --15716425660.B3CBC0.9711-- --===============0987042163== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: base64 Content-Disposition: inline X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KZHJpLWRldmVs IG1haWxpbmcgbGlzdApkcmktZGV2ZWxAbGlzdHMuZnJlZWRlc2t0b3Aub3JnCmh0dHBzOi8vbGlz dHMuZnJlZWRlc2t0b3Aub3JnL21haWxtYW4vbGlzdGluZm8vZHJpLWRldmVs --===============0987042163==--