Attachment 144757 Details for Bug 111103 – [Patch] to compile amdgpu-dkms 19.20 on debian stretch & buster

[patch] [Patch] to compile amdgpu-dkms 19.20 on debian stretch & buster

01-amdgpu_19.20_on_linux_4.19-v1.patch (text/plain), 3.36 KB, created by Fab Stz on 2019-07-11 06:55:55 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Fab Stz

Created: 2019-07-11 06:55:55 UTC

Size: 3.36 KB

patch

obsolete

>diff -Naur amdgpu-19.20-812932.bak/amd/amdgpu/amdgpu_mn.c amdgpu-19.20-812932/amd/amdgpu/amdgpu_mn.c
>--- amdgpu-19.20-812932.bak/amd/amdgpu/amdgpu_mn.c	2019-05-10 19:57:47.000000000 +0200
>+++ amdgpu-19.20-812932/amd/amdgpu/amdgpu_mn.c	2019-07-11 08:30:51.916515905 +0200
>@@ -257,7 +257,8 @@
> 	}
> }
> 
>-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 19, 0)
>+//#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 19, 0)
>+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 0, 0)
> /**
>  * amdgpu_mn_invalidate_range_start_gfx - callback to notify about mm change
>  *
>@@ -365,10 +366,11 @@
>  * Block for operations on BOs to finish and mark pages as accessed and
>  * potentially dirty.
>  */
>-static void amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn,
>+static int amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn,
> 						 struct mm_struct *mm,
> 						 unsigned long start,
>-						 unsigned long end)
>+						 unsigned long end,
>+						 bool blockable)
> {
> 	struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
> 	struct interval_tree_node *it;
>@@ -376,17 +378,27 @@
> 	/* notification is exclusive, but interval is inclusive */
> 	end -= 1;
> 
>-	amdgpu_mn_read_lock(amn);
>+	/* TODO we should be able to split locking for interval tree and
>+	 * amdgpu_mn_invalidate_node
>+	 */
>+	if (amdgpu_mn_read_lock(amn, blockable))
>+		return -EAGAIN;
> 
> 	it = interval_tree_iter_first(&amn->objects, start, end);
> 	while (it) {
> 		struct amdgpu_mn_node *node;
> 
>+		if (!blockable) {
>+			amdgpu_mn_read_unlock(amn);
>+			return -EAGAIN;
>+		}
>+
> 		node = container_of(it, struct amdgpu_mn_node, it);
> 		it = interval_tree_iter_next(it, start, end);
> 
> 		amdgpu_mn_invalidate_node(node, start, end);
> 	}
>+	return 0;
> }
> 
> 
>@@ -402,10 +414,11 @@
>  * necessitates evicting all user-mode queues of the process. The BOs
>  * are restorted in amdgpu_mn_invalidate_range_end_hsa.
>  */
>-static void amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn,
>+static int amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn,
> 						 struct mm_struct *mm,
> 						 unsigned long start,
>-						 unsigned long end)
>+						 unsigned long end,
>+						 bool blockable)
> {
> 	struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
> 	struct interval_tree_node *it;
>@@ -413,13 +426,19 @@
> 	/* notification is exclusive, but interval is inclusive */
> 	end -= 1;
> 
>-	amdgpu_mn_read_lock(amn);
>+	if (amdgpu_mn_read_lock(amn, blockable))
>+		return -EAGAIN;
> 
> 	it = interval_tree_iter_first(&amn->objects, start, end);
> 	while (it) {
> 		struct amdgpu_mn_node *node;
> 		struct amdgpu_bo *bo;
> 
>+		if (!blockable) {
>+			amdgpu_mn_read_unlock(amn);
>+			return -EAGAIN;
>+		}
>+
> 		node = container_of(it, struct amdgpu_mn_node, it);
> 		it = interval_tree_iter_next(it, start, end);
> 
>@@ -431,6 +450,7 @@
> 				amdgpu_amdkfd_evict_userptr(mem, mm);
> 		}
> 	}
>+	return 0;
> }
> 
> #endif
>diff -Naur amdgpu-19.20-812932.bak/Makefile amdgpu-19.20-812932/Makefile
>--- amdgpu-19.20-812932.bak/Makefile	2019-05-10 19:57:47.000000000 +0200
>+++ amdgpu-19.20-812932/Makefile	2019-07-11 07:51:58.646571741 +0200
>@@ -33,6 +33,9 @@
> DRM_VER=$(shell sed -n 's/^VERSION = $.*$/\1/p' $(kdir)/Makefile)
> DRM_PATCH=$(shell sed -n 's/^PATCHLEVEL = $.*$/\1/p' $(kdir)/Makefile)
> endif
>+else
>+DRM_VER=$(shell make kernelversion | sed -n 's/$.*$\.$.*$\.$.*$/\1/p')
>+DRM_PATCH=$(shell make kernelversion | sed -n 's/$.*$\.$.*$\.$.*$/\2/p')
> endif
> 
> subdir-ccflags-y += \

Actions: View | Splinter Review

Attachments on bug 111103: 144757 | 144809

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.