On 2021-06-08 18:18, Eero Lehtinen wrote:
This patch removes Panfrost spamming messages to syslog that causes a poor performance and crashes of the Xfce desktop with a Amlogic S912 TV box. See the old bug in: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3143
There is certainly an argument for demoting the shrinker messages to a debug feature, if we're happy that our memory management is mature enough that memory pressure issues are no longer a common thing to need to spot easily (and especially since msm, from whence that was copied, has now taken a similar route with commit fdf38426cda6). The rest, however, represent genuine error conditions, so hiding them is nonsensical. If the GPU is faulting or locked up, 99% of the time it's because the userspace driver has done *something* wrong (for incredibly many values of "something"), but there's still the possibility of things going wrong in the kernel too. Either way, the bugs only have a hope of getting fixed if we know they're happening.
Taking the "this is fine" approach of wilfully ignoring significant errors because your particular use-case happens to be able to limp along in spite of them might be OK as a local hack, but it is not reasonable for the mainline kernel.
Thanks, Robin.
Signed-off-by: Eero Lehtinen <debiangamer2 at gmail.com>
--- a/drivers/gpu/drm/panfrost/panfrost_mmu.c +++ b/drivers/gpu/drm/panfrost/panfrost_mmu.c @@ -36,8 +36,8 @@ ret = readl_relaxed_poll_timeout_atomic(pfdev->iomem + AS_STATUS(as_nr), val, !(val & AS_STATUS_AS_ACTIVE), 10, 1000);
- if (ret)
dev_err(pfdev->dev, "AS_ACTIVE bit stuck\n");
//if (ret)
// dev_err(pfdev->dev, "AS_ACTIVE bit stuck\n");
return ret; }
--- a/drivers/gpu/drm/panfrost/panfrost_mmu.c +++ b/drivers/gpu/drm/panfrost/panfrost_mmu.c @@ -292,8 +292,8 @@ prot |= IOMMU_NOEXEC;
sgt = drm_gem_shmem_get_pages_sgt(obj);
- if (WARN_ON(IS_ERR(sgt)))
return PTR_ERR(sgt);
//if (WARN_ON(IS_ERR(sgt)))
// return PTR_ERR(sgt);
mmu_map_sg(pfdev, mapping->mmu, mapping->mmnode.start << PAGE_SHIFT, prot, sgt);
--- a/drivers/gpu/drm/panfrost/panfrost_mmu.c +++ b/drivers/gpu/drm/panfrost/panfrost_mmu.c @@ -660,7 +660,7 @@ ret = -1; if ((status & mask) == BIT(as) && (exception_type & 0xF8) == 0xC0) ret = panfrost_mmu_map_fault_addr(pfdev, as, addr);
+#if 0 if (ret) /* terminal fault, print info about the fault */ dev_err(pfdev->dev, @@ -678,7 +678,7 @@ exception_type, panfrost_exception_name(pfdev, exception_type), access_type, access_type_name(pfdev, fault_status), source_id);
+#endif status &= ~mask;
/* If we received new MMU interrupts, process them before returning. */--- a/drivers/gpu/drm/panfrost/panfrost_gem_shrinker.c +++ b/drivers/gpu/drm/panfrost/panfrost_gem_shrinker.c @@ -85,8 +85,8 @@
mutex_unlock(&pfdev->shrinker_lock);
- if (freed > 0)
pr_info_ratelimited("Purging %lu bytes\n", freed << PAGE_SHIFT);
// if (freed > 0)
// pr_info_ratelimited("Purging %lu bytes\n", freed << PAGE_SHIFT); return freed;}