FAILED: patch "[PATCH] drm/amdkfd: don't allow mapping the MMIO HDP page with large" failed to apply to 5.15-stable tree

FAILED: patch "[PATCH] drm/amdkfd: don't allow mapping the MMIO HDP page with large" failed to apply to 5.15-stable tree

[gregkh]

The patch below does not apply to the 5.15-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable@vger.kernel.org>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-5.15.y
git checkout FETCH_HEAD
git cherry-pick -x 24e82654e98e96cece5d8b919c522054456eeec6
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<stable@vger.kernel.org>' --in-reply-to '2024081202-hastily-panic-ee65@gregkh' --subject-prefix 'PATCH 5.15.y' HEAD^..

Possible dependencies:

24e82654e98e ("drm/amdkfd: don't allow mapping the MMIO HDP page with large pages")
b38c074b2b07 ("drm/amdkfd: CRIU Refactor restore BO function")
67a359d85ec2 ("drm/amdkfd: CRIU remove sync and TLB flush on restore")
22804e03f7a5 ("drm/amdkfd: Fix criu_restore_bo error handling")
d8a25e485857 ("drm/amdkfd: fix loop error handling")
e5af61ffaaef ("drm/amdkfd: CRIU fix a NULL vs IS_ERR() check")
be072b06c739 ("drm/amdkfd: CRIU export BOs as prime dmabuf objects")
bef153b70c6e ("drm/amdkfd: CRIU implement gpu_id remapping")
40e8a766a761 ("drm/amdkfd: CRIU checkpoint and restore events")
42c6c48214b7 ("drm/amdkfd: CRIU checkpoint and restore queue mqds")
2485c12c980a ("drm/amdkfd: CRIU restore sdma id for queues")
8668dfc30d3e ("drm/amdkfd: CRIU restore queue ids")
626f7b3190b4 ("drm/amdkfd: CRIU add queues support")
cd9f79103003 ("drm/amdkfd: CRIU Implement KFD unpause operation")
011bbb03024f ("drm/amdkfd: CRIU Implement KFD resume ioctl")
73fa13b6a511 ("drm/amdkfd: CRIU Implement KFD restore ioctl")
5ccbb057c0a1 ("drm/amdkfd: CRIU Implement KFD checkpoint ioctl")
f185381b6481 ("drm/amdkfd: CRIU Implement KFD process_info ioctl")
3698807094ec ("drm/amdkfd: CRIU Introduce Checkpoint-Restore APIs")
f61c40c0757a ("drm/amdkfd: enable heavy-weight TLB flush on Arcturus")

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 24e82654e98e96cece5d8b919c522054456eeec6 Mon Sep 17 00:00:00 2001
From: Alex Deucher <alexander.deucher@amd.com>
Date: Sun, 14 Apr 2024 13:06:39 -0400
Subject: [PATCH] drm/amdkfd: don't allow mapping the MMIO HDP page with large
 pages

We don't get the right offset in that case.  The GPU has
an unused 4K area of the register BAR space into which you can
remap registers.  We remap the HDP flush registers into this
space to allow userspace (CPU or GPU) to flush the HDP when it
updates VRAM.  However, on systems with >4K pages, we end up
exposing PAGE_SIZE of MMIO space.

Fixes: d8e408a82704 ("drm/amdkfd: Expose HDP registers to user space")
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 6b713fb0b818..fdf171ad4a3c 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1144,7 +1144,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 			goto err_unlock;
 		}
 		offset = dev->adev->rmmio_remap.bus_addr;
-		if (!offset) {
+		if (!offset || (PAGE_SIZE > 4096)) {
 			err = -ENOMEM;
 			goto err_unlock;
 		}
@@ -2312,7 +2312,7 @@ static int criu_restore_memory_of_gpu(struct kfd_process_device *pdd,
 			return -EINVAL;
 		}
 		offset = pdd->dev->adev->rmmio_remap.bus_addr;
-		if (!offset) {
+		if (!offset || (PAGE_SIZE > 4096)) {
 			pr_err("amdgpu_amdkfd_get_mmio_remap_phys_addr failed\n");
 			return -ENOMEM;
 		}
@@ -3354,6 +3354,9 @@ static int kfd_mmio_mmap(struct kfd_node *dev, struct kfd_process *process,
 	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
 		return -EINVAL;
 
+	if (PAGE_SIZE > 4096)
+		return -EINVAL;
+
 	address = dev->adev->rmmio_remap.bus_addr;
 
 	vm_flags_set(vma, VM_IO | VM_DONTCOPY | VM_DONTEXPAND | VM_NORESERVE |

[PATCH 5.15.y] drm/amdkfd: don't allow mapping the MMIO HDP page with large pages

[Felix Kuehling]

From: Alex Deucher <alexander.deucher@amd.com>

We don't get the right offset in that case.  The GPU has
an unused 4K area of the register BAR space into which you can
remap registers.  We remap the HDP flush registers into this
space to allow userspace (CPU or GPU) to flush the HDP when it
updates VRAM.  However, on systems with >4K pages, we end up
exposing PAGE_SIZE of MMIO space.

Fixes: d8e408a82704 ("drm/amdkfd: Expose HDP registers to user space")
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
(cherry picked from commit 24e82654e98e96cece5d8b919c522054456eeec6)
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
---
 drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 88f9e1aa51f8..34c466e8eee9 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1290,7 +1290,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 			goto err_unlock;
 		}
 		offset = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd);
-		if (!offset) {
+		if (!offset || (PAGE_SIZE > 4096)) {
 			err = -ENOMEM;
 			goto err_unlock;
 		}
@@ -2029,6 +2029,9 @@ static int kfd_mmio_mmap(struct kfd_dev *dev, struct kfd_process *process,
 	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
 		return -EINVAL;
 
+	if (PAGE_SIZE > 4096)
+		return -EINVAL;
+
 	address = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd);
 
 	vma->vm_flags |= VM_IO | VM_DONTCOPY | VM_DONTEXPAND | VM_NORESERVE |
-- 
2.34.1

Re: [PATCH 5.15.y] drm/amdkfd: don't allow mapping the MMIO HDP page with large pages

[Greg KH]

On Thu, Aug 15, 2024 at 03:46:29PM -0400, Felix Kuehling wrote:
> From: Alex Deucher <alexander.deucher@amd.com>
> 
> We don't get the right offset in that case.  The GPU has
> an unused 4K area of the register BAR space into which you can
> remap registers.  We remap the HDP flush registers into this
> space to allow userspace (CPU or GPU) to flush the HDP when it
> updates VRAM.  However, on systems with >4K pages, we end up
> exposing PAGE_SIZE of MMIO space.
> 
> Fixes: d8e408a82704 ("drm/amdkfd: Expose HDP registers to user space")
> Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> Cc: stable@vger.kernel.org
> (cherry picked from commit 24e82654e98e96cece5d8b919c522054456eeec6)

Wrong git id, this should be be4a2a81b6b9 ("drm/amdkfd: don't allow
mapping the MMIO HDP page with large pages"), right?

{sigh}