[RFC drm-misc-next PATCH 1/1] drm/amdgpu: clean up bo in vce and vcn test
Christian König
christian.koenig at amd.com
Tue Dec 8 14:45:43 UTC 2020
Yes, correct.
You could add an amdgpu_bo_free_reserved() function, but we only have
this one case for that so I think it's probably not worth it.
Just one more comment below.
Am 08.12.20 um 15:42 schrieb Nirmoy:
> I think I know why I needed to keep that amdgpu_bo_unreserve() before
> calling amdgpu_bo_free_kernel().
>
> amdgpu_bo_free_kernel() --> amdgpu_bo_reserve()->ttm_bo_reserve()-->
> dma_resv_lock(bo->base.resv, ticket).
>
> amdgpu_bo_create_reserved() already locked that dma_resv by calling
> amdgpu_bo_reserve(). So "amdgpu_bo_create_reserved();
> amdgpu_bo_free_kernel();" will lead to deadlock.
>
>
> I wonder if we should have a separate API to clean BO created by
> amdgpu_bo_create_reserved().
>
>
> Regards,
>
> Nirmoy
>
> On 12/8/20 3:23 PM, Nirmoy Das wrote:
>> BO created with amdgpu_bo_create_reserved wasn't clean
>> properly before which causes:
>>
>> [ 21.056218] WARNING: CPU: 0 PID: 7 at
>> drivers/gpu/drm/ttm/ttm_bo.c:518 ttm_bo_release+0x2bf/0x310 [ttm]
>> [ 21.056219] Modules linked in: amdgpu(E) iommu_v2(E) gpu_sched(E)
>> drm_ttm_helper(E) ttm(E) drm_kms_helper(E) syscopyarea(E)
>> sysfillrect(E) sysimgblt(E) fb_sys_fops(E) cec(E) rc_core(E)
>> rfcomm(E) af_packet(E) cmac(E) algif_hash(E) algif_skcipher(E)
>> af_alg(E) xt_CHECKSUM(E) xt_MASQUERADE(E) nf_nat_tftp(E)
>> nf_conntrack_tftp(E) xt_CT(E) bridge(E) stp(E) llc(E) ip6t_REJECT(E)
>> nf_reject_ipv6(E) ip6t_rpfilter(E) xt_tcpudp(E) ipt_REJECT(E)
>> nf_reject_ipv4(E) xt_conntrack(E) ebtable_nat(E) ebtable_broute(E)
>> ip6table_nat(E) ip6table_mangle(E) ip6table_raw(E)
>> ip6table_security(E) iptable_nat(E) nf_nat(E) nf_conntrack(E)
>> nf_defrag_ipv6(E) nf_defrag_ipv4(E) iptable_mangle(E) iptable_raw(E)
>> iptable_security(E) ip_set(E) nfnetlink(E) ebtable_filter(E)
>> ebtables(E) ip6table_filter(E) ip6_tables(E) iptable_filter(E)
>> ip_tables(E) x_tables(E) bpfilter(E) joydev(E) hid_generic(E)
>> usbhid(E) bnep(E) btusb(E) btrtl(E) btbcm(E) btintel(E) bluetooth(E)
>> ecdh_generic(E) ecc(E) rfkill(E) dmi_sysfs(E) msr(E) wmi_bmof(E)
>> [ 21.056266] mxm_wmi(E) snd_hda_codec_realtek(E)
>> snd_hda_codec_generic(E) ledtrig_audio(E) snd_hda_codec_hdmi(E)
>> kvm_amd(E) snd_hda_intel(E) kvm(E) snd_intel_dspcfg(E)
>> snd_hda_codec(E) snd_hwdep(E) snd_hda_core(E) irqbypass(E) snd_pcm(E)
>> snd_timer(E) snd(E) soundcore(E) igb(E) sp5100_tco(E) pcspkr(E)
>> i2c_piix4(E) k10temp(E) i2c_algo_bit(E) dca(E) wmi(E)
>> tiny_power_button(E) gpio_amdpt(E) gpio_generic(E) acpi_cpufreq(E)
>> button(E) drm(E) crct10dif_pclmul(E) crc32_pclmul(E)
>> ghash_clmulni_intel(E) aesni_intel(E) glue_helper(E) crypto_simd(E)
>> cryptd(E) xhci_pci(E) nvme(E) xhci_pci_renesas(E) nvme_core(E)
>> xhci_hcd(E) ccp(E) usbcore(E) btrfs(E) blake2b_generic(E)
>> libcrc32c(E) crc32c_intel(E) xor(E) raid6_pq(E) sg(E) dm_multipath(E)
>> dm_mod(E) scsi_dh_rdac(E) scsi_dh_emc(E) scsi_dh_alua(E)
>> [ 21.056307] CPU: 0 PID: 7 Comm: kworker/0:1 Tainted: G
>> E 5.10.0-rc3-1-default+ #45
>> [ 21.056309] Hardware name: Gigabyte Technology Co., Ltd. X399
>> DESIGNARE EX/X399 DESIGNARE EX-CF, BIOS F12i 09/24/2019
>> [ 21.056409] Workqueue: events
>> amdgpu_device_delayed_init_work_handler [amdgpu]
>> [ 21.056415] RIP: 0010:ttm_bo_release+0x2bf/0x310 [ttm]
>> [ 21.056418] Code: e9 a1 fd ff ff e8 b1 00 f0 db e9 d2 fd ff ff 49
>> 8b 7e 88 b9 4c 1d 00 00 31 d2 be 01 00 00 00 e8 87 27 f0 db 49 8b 46
>> d8 eb 9e <0f> 0b 41 c7 86 9c 00 00 00 00 00 00 00 4c 89 ef e8 9c ef
>> ff ff 49
>> [ 21.056419] RSP: 0018:ffff9cca00107d48 EFLAGS: 00010202
>> [ 21.056421] RAX: 0000000000000001 RBX: 0000000000000000 RCX:
>> 000000001ec00000
>> [ 21.056423] RDX: 0000000000000001 RSI: 0000000000000246 RDI:
>> ffffffffc0e61b28
>> [ 21.056424] RBP: ffff90fa735e55b8 R08: ffff90fa731e8db8 R09:
>> 0000000000000000
>> [ 21.056425] R10: ffff90fa7335e000 R11: ffff90fa7335e000 R12:
>> ffffffffc0e61b28
>> [ 21.056426] R13: ffff90fa465f3c58 R14: ffff90fa465f3dc8 R15:
>> ffff90fa5622d600
>> [ 21.056427] FS: 0000000000000000(0000) GS:ffff91019ec00000(0000)
>> knlGS:0000000000000000
>> [ 21.056428] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>> [ 21.056429] CR2: 00007f1150baeff8 CR3: 0000000116c70000 CR4:
>> 00000000003506f0
>> [ 21.056430] Call Trace:
>> [ 21.056525] amdgpu_bo_unref+0x1a/0x30 [amdgpu]
>> [ 21.056635] amdgpu_vcn_dec_send_msg+0x1b2/0x270 [amdgpu]
>> [ 21.056740] amdgpu_vcn_dec_get_create_msg.constprop.0+0xd8/0x100
>> [amdgpu]
>> [ 21.056843] amdgpu_vcn_dec_ring_test_ib+0x27/0x180 [amdgpu]
>> [ 21.056936] amdgpu_ib_ring_tests+0xf1/0x150 [amdgpu]
>> [ 21.057024] amdgpu_device_delayed_init_work_handler+0x11/0x30
>> [amdgpu]
>> [ 21.057030] process_one_work+0x1df/0x370
>> [ 21.057033] worker_thread+0x46/0x340
>> [ 21.057034] ? process_one_work+0x370/0x370
>> [ 21.057037] kthread+0x11b/0x140
>> [ 21.057039] ? __kthread_bind_mask+0x60/0x60
>> [ 21.057043] ret_from_fork+0x22/0x30
>>
>> Signed-off-by: Nirmoy Das <nirmoy.das at amd.com>
>> ---
>>
>> I had to keep amdgpu_bo_unreserve() before calling
>> amdgpu_bo_free_kernel()
>> or else amdgpu doesn't respond after loading. Is there any better
>> solution ?
>>
>> drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 2 +-
>> drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 16 ++++++++++------
>> 2 files changed, 11 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
>> index ecaa2d7483b2..78a4dd9bf11f 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
>> @@ -1151,6 +1151,6 @@ int amdgpu_vce_ring_test_ib(struct amdgpu_ring
>> *ring, long timeout)
>> error:
>> dma_fence_put(fence);
>> amdgpu_bo_unreserve(bo);
>> - amdgpu_bo_unref(&bo);
>> + amdgpu_bo_free_kernel(&bo, NULL, NULL);
>> return r;
>> }
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> index 7e19a6656715..dfcdd38ff9c2 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
>> @@ -491,8 +491,6 @@ static int amdgpu_vcn_dec_send_msg(struct
>> amdgpu_ring *ring,
>> goto err_free;
>>
>> amdgpu_bo_fence(bo, f, false);
>> - amdgpu_bo_unreserve(bo);
>> - amdgpu_bo_unref(&bo);
>>
>> if (fence)
>> *fence = dma_fence_get(f);
>> @@ -504,8 +502,6 @@ static int amdgpu_vcn_dec_send_msg(struct
>> amdgpu_ring *ring,
>> amdgpu_job_free(job);
>>
>> err:
>> - amdgpu_bo_unreserve(bo);
>> - amdgpu_bo_unref(&bo);
>> return r;
>> }
>>
>> @@ -540,7 +536,11 @@ static int amdgpu_vcn_dec_get_create_msg(struct
>> amdgpu_ring *ring, uint32_t hand
>> for (i = 14; i < 1024; ++i)
>> msg[i] = cpu_to_le32(0x0);
>>
>> - return amdgpu_vcn_dec_send_msg(ring, bo, fence);
>> + r = amdgpu_vcn_dec_send_msg(ring, bo, fence);
>> + amdgpu_bo_unreserve(bo);
>> + amdgpu_bo_free_kernel(&bo, NULL, (void **)&msg);
Why did you moved that here? As far as I see you just need to replace
the amdgpu_bo_unref() above with amdgpu_bo_free_kernel().
Christian.
>> +
>> + return r;
>> }
>>
>> static int amdgpu_vcn_dec_get_destroy_msg(struct amdgpu_ring *ring,
>> uint32_t handle,
>> @@ -566,7 +566,11 @@ static int amdgpu_vcn_dec_get_destroy_msg(struct
>> amdgpu_ring *ring, uint32_t han
>> for (i = 6; i < 1024; ++i)
>> msg[i] = cpu_to_le32(0x0);
>>
>> - return amdgpu_vcn_dec_send_msg(ring, bo, fence);
>> + r = amdgpu_vcn_dec_send_msg(ring, bo, fence);
>> + amdgpu_bo_unreserve(bo);
>> + amdgpu_bo_free_kernel(&bo, NULL, (void **)&msg);
>> +
>> + return r;
>> }
>>
>> int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long
>> timeout)
>> --
>> 2.29.2
>>
>
More information about the amd-gfx
mailing list