Re: [PATCH] drm/amdkfd: Only apply heavy-weight TLB flush on Aldebaran

[Eric Huang <jinhuieric.huang@xxxxxxx>]

Hi Felix,

I was not able to reproduce the VM fault issue of SWDEV-249241, which is 
the only regression reported on MI200. So the patch is valid to review. 
Please take a look.

Thanks,
Eric

On 2021-07-09 1:45 a.m., Chen, Guchun wrote:
> [Public]
>
> Original patch will cause regressions on Aldebaran as well, so this workaround is still invalid.
>
> Regards,
> Guchun
>
> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@xxxxxxxxxxxxxxxxxxxxx> On Behalf Of Eric Huang
> Sent: Friday, July 9, 2021 3:54 AM
> To: amd-gfx@xxxxxxxxxxxxxxxxxxxxx
> Cc: Huang, JinHuiEric <JinHuiEric.Huang@xxxxxxx>; Kuehling, Felix <Felix.Kuehling@xxxxxxx>
> Subject: [PATCH] drm/amdkfd: Only apply heavy-weight TLB flush on Aldebaran
>
> It is to workaround HW bug on other Asics and based on reverting two commits:
>    drm/amdkfd: Add heavy-weight TLB flush after unmapping
>    drm/amdkfd: Add memory sync before TLB flush on unmap
>
> Signed-off-by: Eric Huang <jinhuieric.huang@xxxxxxx>
> ---
>   drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 37 +++++++++++++-----------
>   1 file changed, 20 insertions(+), 17 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> index ebb4872c5a9d..5f2655cf0162 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
> @@ -1773,26 +1773,29 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep,
>   	}
>   	mutex_unlock(&p->mutex);
>   
> -	err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd, (struct kgd_mem *) mem, true);
> -	if (err) {
> -		pr_debug("Sync memory failed, wait interrupted by user signal\n");
> -		goto sync_memory_failed;
> -	}
> +	if (dev->device_info->asic_family == CHIP_ALDEBARAN) {
> +		err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd,
> +				(struct kgd_mem *) mem, true);
>   
> -	/* Flush TLBs after waiting for the page table updates to complete */
> -	for (i = 0; i < args->n_devices; i++) {
> -		peer = kfd_device_by_id(devices_arr[i]);
> -		if (WARN_ON_ONCE(!peer))
> -			continue;
> -		peer_pdd = kfd_get_process_device_data(peer, p);
> -		if (WARN_ON_ONCE(!peer_pdd))
> -			continue;
> -		if (!amdgpu_read_lock(peer->ddev, true)) {
> -			kfd_flush_tlb(peer_pdd, TLB_FLUSH_HEAVYWEIGHT);
> -			amdgpu_read_unlock(peer->ddev);
> +		if (err) {
> +			pr_debug("Sync memory failed, wait interrupted by user signal\n");
> +			goto sync_memory_failed;
>   		}
> -	}
>   
> +		/* Flush TLBs after waiting for the page table updates to complete */
> +		for (i = 0; i < args->n_devices; i++) {
> +			peer = kfd_device_by_id(devices_arr[i]);
> +			if (WARN_ON_ONCE(!peer))
> +				continue;
> +			peer_pdd = kfd_get_process_device_data(peer, p);
> +			if (WARN_ON_ONCE(!peer_pdd))
> +				continue;
> +			if (!amdgpu_read_lock(peer->ddev, true)) {
> +				kfd_flush_tlb(peer_pdd, TLB_FLUSH_HEAVYWEIGHT);
> +				amdgpu_read_unlock(peer->ddev);
> +			}
> +		}
> +	}
>   	kfree(devices_arr);
>   
>   	return 0;
> --
> 2.25.1
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@xxxxxxxxxxxxxxxxxxxxx
> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=04%7C01%7Cguchun.chen%40amd.com%7C69113cf367eb450a8f8808d9424a23fe%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637613708477013366%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&amp;sdata=0ESYvG5kCSJaFT9dR4jW5VacL8x7TghGw1aKWTRa9R4%3D&amp;reserved=0

_______________________________________________
amd-gfx mailing list
amd-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/amd-gfx