[PATCH] gpu: host1x: Do not leak BO's phys address to userspace

[PATCH] gpu: host1x: Do not leak BO's phys address to userspace

[Dmitry Osipenko]

Do gathers coping before patching them, so the original gathers are left
untouched. That's not as bad as leaking a kernel addresses, but still
doesn't feel right.

Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
---
 drivers/gpu/host1x/job.c | 46 ++++++++++++++++++++++++++++++----------------
 1 file changed, 30 insertions(+), 16 deletions(-)

diff --git a/drivers/gpu/host1x/job.c b/drivers/gpu/host1x/job.c
index d9933828fe87..8f0d43b8f9a6 100644
--- a/drivers/gpu/host1x/job.c
+++ b/drivers/gpu/host1x/job.c
@@ -137,8 +137,9 @@ static void host1x_syncpt_patch_offset(struct host1x_syncpt *sp,
  * avoid a wrap condition in the HW).
  */
 static int do_waitchks(struct host1x_job *job, struct host1x *host,
-		       struct host1x_bo *patch)
+		       struct host1x_job_gather *g)
 {
+	struct host1x_bo *patch = g->bo;
 	int i;
 
 	/* compare syncpt vs wait threshold */
@@ -165,7 +166,8 @@ static int do_waitchks(struct host1x_job *job, struct host1x *host,
 				wait->syncpt_id, sp->name, wait->thresh,
 				host1x_syncpt_read_min(sp));
 
-			host1x_syncpt_patch_offset(sp, patch, wait->offset);
+			host1x_syncpt_patch_offset(sp, patch,
+						   g->offset + wait->offset);
 		}
 
 		wait->bo = NULL;
@@ -269,11 +271,12 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 	return err;
 }
 
-static int do_relocs(struct host1x_job *job, struct host1x_bo *cmdbuf)
+static int do_relocs(struct host1x_job *job, struct host1x_job_gather *g)
 {
 	int i = 0;
 	u32 last_page = ~0;
 	void *cmdbuf_page_addr = NULL;
+	struct host1x_bo *cmdbuf = g->bo;
 
 	/* pin & patch the relocs for one gather */
 	for (i = 0; i < job->num_relocs; i++) {
@@ -286,7 +289,8 @@ static int do_relocs(struct host1x_job *job, struct host1x_bo *cmdbuf)
 		if (cmdbuf != reloc->cmdbuf.bo)
 			continue;
 
-		if (last_page != reloc->cmdbuf.offset >> PAGE_SHIFT) {
+		if (!IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL) &&
+		    last_page != reloc->cmdbuf.offset >> PAGE_SHIFT) {
 			if (cmdbuf_page_addr)
 				host1x_bo_kunmap(cmdbuf, last_page,
 						 cmdbuf_page_addr);
@@ -301,11 +305,20 @@ static int do_relocs(struct host1x_job *job, struct host1x_bo *cmdbuf)
 			}
 		}
 
+		if (IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL)) {
+			cmdbuf_page_addr = job->gather_copy_mapped;
+			cmdbuf_page_addr += g->offset;
+		}
+
 		target = cmdbuf_page_addr + (reloc->cmdbuf.offset & ~PAGE_MASK);
+
+		if (IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL))
+			target += (reloc->cmdbuf.offset & PAGE_MASK) >> 2;
+
 		*target = reloc_addr;
 	}
 
-	if (cmdbuf_page_addr)
+	if (!IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL) && cmdbuf_page_addr)
 		host1x_bo_kunmap(cmdbuf, last_page, cmdbuf_page_addr);
 
 	return 0;
@@ -573,6 +586,14 @@ int host1x_job_pin(struct host1x_job *job, struct device *dev)
 	if (err)
 		goto out;
 
+	if (IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL) && !err) {
+		err = copy_gathers(job, dev);
+		if (err) {
+			host1x_job_unpin(job);
+			return err;
+		}
+	}
+
 	/* patch gathers */
 	for (i = 0; i < job->num_gathers; i++) {
 		struct host1x_job_gather *g = &job->gathers[i];
@@ -581,7 +602,8 @@ int host1x_job_pin(struct host1x_job *job, struct device *dev)
 		if (g->handled)
 			continue;
 
-		g->base = job->gather_addr_phys[i];
+		if (!IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL))
+			g->base = job->gather_addr_phys[i];
 
 		for (j = i + 1; j < job->num_gathers; j++) {
 			if (job->gathers[j].bo == g->bo) {
@@ -590,23 +612,15 @@ int host1x_job_pin(struct host1x_job *job, struct device *dev)
 			}
 		}
 
-		err = do_relocs(job, g->bo);
+		err = do_relocs(job, g);
 		if (err)
 			break;
 
-		err = do_waitchks(job, host, g->bo);
+		err = do_waitchks(job, host, g);
 		if (err)
 			break;
 	}
 
-	if (IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL) && !err) {
-		err = copy_gathers(job, dev);
-		if (err) {
-			host1x_job_unpin(job);
-			return err;
-		}
-	}
-
 out:
 	wmb();
 
-- 
2.13.0

[PATCH] drm/tegra: Check whether page belongs to BO in tegra_bo_kmap()

[Dmitry Osipenko]

This fixes an OOPS in case of out-of-bounds accessing of a kmap'ed commands
buffer CMA while patching relocations in do_relocs().

Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
---
 drivers/gpu/drm/tegra/gem.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/gpu/drm/tegra/gem.c b/drivers/gpu/drm/tegra/gem.c
index 424569b53e57..b76d7ac75696 100644
--- a/drivers/gpu/drm/tegra/gem.c
+++ b/drivers/gpu/drm/tegra/gem.c
@@ -74,6 +74,9 @@ static void *tegra_bo_kmap(struct host1x_bo *bo, unsigned int page)
 {
 	struct tegra_bo *obj = host1x_to_tegra_bo(bo);
 
+	if (page * PAGE_SIZE > obj->gem.size)
+		return NULL;
+
 	if (obj->vaddr)
 		return obj->vaddr + page * PAGE_SIZE;
 	else if (obj->gem.import_attach)
-- 
2.13.0

[PATCH] drm/tegra: Check offsets of a submitted command buffer and of relocations

[Dmitry Osipenko]

If commands buffer claims a number of words that is higher than its BO can
fit, a kernel OOPS will be fired on the out-of-bounds BO access. This was
triggered by an opentegra Xorg driver that erroneously pushed too many
commands to the pushbuf. The CMDA commands buffer address is 4 bytes
aligned, so check the alignment as well.

Add a sanity check for the relocations in a same way.

[   46.829393] Unable to handle kernel paging request at virtual address f09b2000
...
[<c04a3ba4>] (host1x_job_pin) from [<c04dfcd0>] (tegra_drm_submit+0x474/0x510)
[<c04dfcd0>] (tegra_drm_submit) from [<c04deea0>] (tegra_submit+0x50/0x6c)
[<c04deea0>] (tegra_submit) from [<c04c07c0>] (drm_ioctl+0x1e4/0x3ec)
[<c04c07c0>] (drm_ioctl) from [<c02541a0>] (do_vfs_ioctl+0x9c/0x8e4)
[<c02541a0>] (do_vfs_ioctl) from [<c0254a1c>] (SyS_ioctl+0x34/0x5c)
[<c0254a1c>] (SyS_ioctl) from [<c0107640>] (ret_fast_syscall+0x0/0x3c)

Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Reviewed-by: Erik Faye-Lund <kusmabite-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
---
 drivers/gpu/drm/tegra/drm.c | 30 ++++++++++++++++++++++++++++++
 drivers/gpu/drm/tegra/gem.c |  5 -----
 drivers/gpu/drm/tegra/gem.h |  5 +++++
 3 files changed, 35 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/tegra/drm.c b/drivers/gpu/drm/tegra/drm.c
index 768750226452..c5844a065681 100644
--- a/drivers/gpu/drm/tegra/drm.c
+++ b/drivers/gpu/drm/tegra/drm.c
@@ -362,6 +362,8 @@ int tegra_drm_submit(struct tegra_drm_context *context,
 	while (num_cmdbufs) {
 		struct drm_tegra_cmdbuf cmdbuf;
 		struct host1x_bo *bo;
+		struct tegra_bo *obj;
+		u64 offset;
 
 		if (copy_from_user(&cmdbuf, cmdbufs, sizeof(cmdbuf))) {
 			err = -EFAULT;
@@ -374,6 +376,14 @@ int tegra_drm_submit(struct tegra_drm_context *context,
 			goto fail;
 		}
 
+		offset = (u64)cmdbuf.offset + (u64)cmdbuf.words * sizeof(u32);
+		obj = host1x_to_tegra_bo(bo);
+
+		if (offset & 3 || offset > obj->gem.size) {
+			err = -EINVAL;
+			goto fail;
+		}
+
 		host1x_job_add_gather(job, bo, cmdbuf.words, cmdbuf.offset);
 		num_cmdbufs--;
 		cmdbufs++;
@@ -381,11 +391,31 @@ int tegra_drm_submit(struct tegra_drm_context *context,
 
 	/* copy and resolve relocations from submit */
 	while (num_relocs--) {
+		struct host1x_reloc *reloc;
+		struct tegra_bo *obj;
+
 		err = host1x_reloc_copy_from_user(&job->relocarray[num_relocs],
 						  &relocs[num_relocs], drm,
 						  file);
 		if (err < 0)
 			goto fail;
+
+		reloc = &job->relocarray[num_relocs];
+		obj = host1x_to_tegra_bo(reloc->cmdbuf.bo);
+
+		if (reloc->cmdbuf.offset & 3 ||
+		    reloc->cmdbuf.offset > obj->gem.size) {
+			err = -EINVAL;
+			goto fail;
+		}
+
+		obj = host1x_to_tegra_bo(reloc->target.bo);
+
+		if (reloc->target.offset & 3 ||
+		    reloc->target.offset > obj->gem.size) {
+			err = -EINVAL;
+			goto fail;
+		}
 	}
 
 	if (copy_from_user(job->waitchk, waitchks,
diff --git a/drivers/gpu/drm/tegra/gem.c b/drivers/gpu/drm/tegra/gem.c
index b76d7ac75696..a0ff30c01ac1 100644
--- a/drivers/gpu/drm/tegra/gem.c
+++ b/drivers/gpu/drm/tegra/gem.c
@@ -20,11 +20,6 @@
 #include "drm.h"
 #include "gem.h"
 
-static inline struct tegra_bo *host1x_to_tegra_bo(struct host1x_bo *bo)
-{
-	return container_of(bo, struct tegra_bo, base);
-}
-
 static void tegra_bo_put(struct host1x_bo *bo)
 {
 	struct tegra_bo *obj = host1x_to_tegra_bo(bo);
diff --git a/drivers/gpu/drm/tegra/gem.h b/drivers/gpu/drm/tegra/gem.h
index 6c5f12ac0087..8b32a6fd586d 100644
--- a/drivers/gpu/drm/tegra/gem.h
+++ b/drivers/gpu/drm/tegra/gem.h
@@ -52,6 +52,11 @@ static inline struct tegra_bo *to_tegra_bo(struct drm_gem_object *gem)
 	return container_of(gem, struct tegra_bo, gem);
 }
 
+static inline struct tegra_bo *host1x_to_tegra_bo(struct host1x_bo *bo)
+{
+	return container_of(bo, struct tegra_bo, base);
+}
+
 struct tegra_bo *tegra_bo_create(struct drm_device *drm, size_t size,
 				 unsigned long flags);
 struct tegra_bo *tegra_bo_create_with_handle(struct drm_file *file,
-- 
2.13.0

[PATCH v2] drm/tegra: Correct idr_alloc() minimum id

[Dmitry Osipenko]

The client ID 0 is reserved by the host1x/cdma to mark the timeout timer
work as already been scheduled and context ID is used as the clients one.
This fixes spurious CDMA timeouts.

Fixes: bdd2f9cd10eb ("drm/tegra: Don't leak kernel pointer to userspace")
Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
---

v2: Changed the commit description, now explains the cause of CDMA timeouts.

 drivers/gpu/drm/tegra/drm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/tegra/drm.c b/drivers/gpu/drm/tegra/drm.c
index c5844a065681..489cb32453f7 100644
--- a/drivers/gpu/drm/tegra/drm.c
+++ b/drivers/gpu/drm/tegra/drm.c
@@ -548,7 +548,7 @@ static int tegra_client_open(struct tegra_drm_file *fpriv,
 	if (err < 0)
 		return err;
 
-	err = idr_alloc(&fpriv->contexts, context, 0, 0, GFP_KERNEL);
+	err = idr_alloc(&fpriv->contexts, context, 1, 0, GFP_KERNEL);
 	if (err < 0) {
 		client->ops->close_channel(context);
 		return err;
-- 
2.13.0

Re: [PATCH] drm/tegra: Check whether page belongs to BO in tegra_bo_kmap()

[Dmitry Osipenko]

On 14.05.2017 23:47, Dmitry Osipenko wrote:
> This fixes an OOPS in case of out-of-bounds accessing of a kmap'ed commands
> buffer CMA while patching relocations in do_relocs().
> 
> Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> ---
>  drivers/gpu/drm/tegra/gem.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/gpu/drm/tegra/gem.c b/drivers/gpu/drm/tegra/gem.c
> index 424569b53e57..b76d7ac75696 100644
> --- a/drivers/gpu/drm/tegra/gem.c
> +++ b/drivers/gpu/drm/tegra/gem.c
> @@ -74,6 +74,9 @@ static void *tegra_bo_kmap(struct host1x_bo *bo, unsigned int page)
>  {
>  	struct tegra_bo *obj = host1x_to_tegra_bo(bo);
>  
> +	if (page * PAGE_SIZE > obj->gem.size)
> +		return NULL;
> +
>  	if (obj->vaddr)
>  		return obj->vaddr + page * PAGE_SIZE;
>  	else if (obj->gem.import_attach)
> 

It should be '>=', I'll wait for the review comments before sending out a new
version of the patch.

-- 
Dmitry

Re: [PATCH] drm/tegra: Check offsets of a submitted command buffer and of relocations

[Dmitry Osipenko]

On 14.05.2017 23:47, Dmitry Osipenko wrote:
>  	/* copy and resolve relocations from submit */
>  	while (num_relocs--) {
> +		struct host1x_reloc *reloc;
> +		struct tegra_bo *obj;
> +
>  		err = host1x_reloc_copy_from_user(&job->relocarray[num_relocs],
>  						  &relocs[num_relocs], drm,
>  						  file);
>  		if (err < 0)
>  			goto fail;
> +
> +		reloc = &job->relocarray[num_relocs];
> +		obj = host1x_to_tegra_bo(reloc->cmdbuf.bo);
> +
> +		if (reloc->cmdbuf.offset & 3 ||
> +		    reloc->cmdbuf.offset > obj->gem.size) {
> +			err = -EINVAL;
> +			goto fail;
> +		}
> +
> +		obj = host1x_to_tegra_bo(reloc->target.bo);
> +
> +		if (reloc->target.offset & 3 ||
> +		    reloc->target.offset > obj->gem.size) {
> +			err = -EINVAL;
> +			goto fail;
> +		}
>  	}

The relocation offsets should be compared as '>=' as well.

-- 
Dmitry

Re: [PATCH] drm/tegra: Check offsets of a submitted command buffer and of relocations

[Mikko Perttunen]

On 14.05.2017 23:47, Dmitry Osipenko wrote:
> If commands buffer claims a number of words that is higher than its BO can
> fit, a kernel OOPS will be fired on the out-of-bounds BO access. This was
> triggered by an opentegra Xorg driver that erroneously pushed too many
> commands to the pushbuf. The CMDA commands buffer address is 4 bytes
> aligned, so check the alignment as well.
>
> Add a sanity check for the relocations in a same way.
>
> [   46.829393] Unable to handle kernel paging request at virtual address f09b2000
> ...
> [<c04a3ba4>] (host1x_job_pin) from [<c04dfcd0>] (tegra_drm_submit+0x474/0x510)
> [<c04dfcd0>] (tegra_drm_submit) from [<c04deea0>] (tegra_submit+0x50/0x6c)
> [<c04deea0>] (tegra_submit) from [<c04c07c0>] (drm_ioctl+0x1e4/0x3ec)
> [<c04c07c0>] (drm_ioctl) from [<c02541a0>] (do_vfs_ioctl+0x9c/0x8e4)
> [<c02541a0>] (do_vfs_ioctl) from [<c0254a1c>] (SyS_ioctl+0x34/0x5c)
> [<c0254a1c>] (SyS_ioctl) from [<c0107640>] (ret_fast_syscall+0x0/0x3c)
>
> Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> Reviewed-by: Erik Faye-Lund <kusmabite-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> ---
>  drivers/gpu/drm/tegra/drm.c | 30 ++++++++++++++++++++++++++++++
>  drivers/gpu/drm/tegra/gem.c |  5 -----
>  drivers/gpu/drm/tegra/gem.h |  5 +++++
>  3 files changed, 35 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/tegra/drm.c b/drivers/gpu/drm/tegra/drm.c
> index 768750226452..c5844a065681 100644
> --- a/drivers/gpu/drm/tegra/drm.c
> +++ b/drivers/gpu/drm/tegra/drm.c
> @@ -362,6 +362,8 @@ int tegra_drm_submit(struct tegra_drm_context *context,
>  	while (num_cmdbufs) {
>  		struct drm_tegra_cmdbuf cmdbuf;
>  		struct host1x_bo *bo;
> +		struct tegra_bo *obj;
> +		u64 offset;
>
>  		if (copy_from_user(&cmdbuf, cmdbufs, sizeof(cmdbuf))) {
>  			err = -EFAULT;
> @@ -374,6 +376,14 @@ int tegra_drm_submit(struct tegra_drm_context *context,
>  			goto fail;
>  		}
>
> +		offset = (u64)cmdbuf.offset + (u64)cmdbuf.words * sizeof(u32);
> +		obj = host1x_to_tegra_bo(bo);
> +
> +		if (offset & 3 || offset > obj->gem.size) {
> +			err = -EINVAL;
> +			goto fail;
> +		}
> +
>  		host1x_job_add_gather(job, bo, cmdbuf.words, cmdbuf.offset);
>  		num_cmdbufs--;
>  		cmdbufs++;
> @@ -381,11 +391,31 @@ int tegra_drm_submit(struct tegra_drm_context *context,
>
>  	/* copy and resolve relocations from submit */
>  	while (num_relocs--) {
> +		struct host1x_reloc *reloc;
> +		struct tegra_bo *obj;
> +
>  		err = host1x_reloc_copy_from_user(&job->relocarray[num_relocs],
>  						  &relocs[num_relocs], drm,
>  						  file);
>  		if (err < 0)
>  			goto fail;
> +
> +		reloc = &job->relocarray[num_relocs];
> +		obj = host1x_to_tegra_bo(reloc->cmdbuf.bo);
> +
> +		if (reloc->cmdbuf.offset & 3 ||
> +		    reloc->cmdbuf.offset > obj->gem.size) {

This could still fail if the bo's size is not divisible by 4, even with 
 >= comparison (we would overwrite the buffer by 1 to 3 bytes). I would 
do the same as in the gather case, i.e. find out the address immediately 
after the write and compare using >. Perhaps add a helper function if it 
makes sense. I also don't think the "& 3" checks are needed.

> +			err = -EINVAL;
> +			goto fail;
> +		}
> +
> +		obj = host1x_to_tegra_bo(reloc->target.bo);
> +
> +		if (reloc->target.offset & 3 ||
> +		    reloc->target.offset > obj->gem.size) {
> +			err = -EINVAL;
> +			goto fail;
> +		}
>  	}
>
>  	if (copy_from_user(job->waitchk, waitchks,
> diff --git a/drivers/gpu/drm/tegra/gem.c b/drivers/gpu/drm/tegra/gem.c
> index b76d7ac75696..a0ff30c01ac1 100644
> --- a/drivers/gpu/drm/tegra/gem.c
> +++ b/drivers/gpu/drm/tegra/gem.c
> @@ -20,11 +20,6 @@
>  #include "drm.h"
>  #include "gem.h"
>
> -static inline struct tegra_bo *host1x_to_tegra_bo(struct host1x_bo *bo)
> -{
> -	return container_of(bo, struct tegra_bo, base);
> -}
> -
>  static void tegra_bo_put(struct host1x_bo *bo)
>  {
>  	struct tegra_bo *obj = host1x_to_tegra_bo(bo);
> diff --git a/drivers/gpu/drm/tegra/gem.h b/drivers/gpu/drm/tegra/gem.h
> index 6c5f12ac0087..8b32a6fd586d 100644
> --- a/drivers/gpu/drm/tegra/gem.h
> +++ b/drivers/gpu/drm/tegra/gem.h
> @@ -52,6 +52,11 @@ static inline struct tegra_bo *to_tegra_bo(struct drm_gem_object *gem)
>  	return container_of(gem, struct tegra_bo, gem);
>  }
>
> +static inline struct tegra_bo *host1x_to_tegra_bo(struct host1x_bo *bo)
> +{
> +	return container_of(bo, struct tegra_bo, base);
> +}
> +
>  struct tegra_bo *tegra_bo_create(struct drm_device *drm, size_t size,
>  				 unsigned long flags);
>  struct tegra_bo *tegra_bo_create_with_handle(struct drm_file *file,
>

Re: [PATCH] drm/tegra: Check offsets of a submitted command buffer and of relocations

[Erik Faye-Lund]

On Tue, May 16, 2017 at 8:56 AM, Mikko Perttunen <cyndis-/1wQRMveznE@public.gmane.org> wrote:
> On 14.05.2017 23:47, Dmitry Osipenko wrote:
>>
>> If commands buffer claims a number of words that is higher than its BO can
>> fit, a kernel OOPS will be fired on the out-of-bounds BO access. This was
>> triggered by an opentegra Xorg driver that erroneously pushed too many
>> commands to the pushbuf. The CMDA commands buffer address is 4 bytes
>> aligned, so check the alignment as well.
>>
>> Add a sanity check for the relocations in a same way.
>>
>> [   46.829393] Unable to handle kernel paging request at virtual address
>> f09b2000
>> ...
>> [<c04a3ba4>] (host1x_job_pin) from [<c04dfcd0>]
>> (tegra_drm_submit+0x474/0x510)
>> [<c04dfcd0>] (tegra_drm_submit) from [<c04deea0>] (tegra_submit+0x50/0x6c)
>> [<c04deea0>] (tegra_submit) from [<c04c07c0>] (drm_ioctl+0x1e4/0x3ec)
>> [<c04c07c0>] (drm_ioctl) from [<c02541a0>] (do_vfs_ioctl+0x9c/0x8e4)
>> [<c02541a0>] (do_vfs_ioctl) from [<c0254a1c>] (SyS_ioctl+0x34/0x5c)
>> [<c0254a1c>] (SyS_ioctl) from [<c0107640>] (ret_fast_syscall+0x0/0x3c)
>>
>> Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>> Reviewed-by: Erik Faye-Lund <kusmabite-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>> ---
>>  drivers/gpu/drm/tegra/drm.c | 30 ++++++++++++++++++++++++++++++
>>  drivers/gpu/drm/tegra/gem.c |  5 -----
>>  drivers/gpu/drm/tegra/gem.h |  5 +++++
>>  3 files changed, 35 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/tegra/drm.c b/drivers/gpu/drm/tegra/drm.c
>> index 768750226452..c5844a065681 100644
>> --- a/drivers/gpu/drm/tegra/drm.c
>> +++ b/drivers/gpu/drm/tegra/drm.c
>> @@ -362,6 +362,8 @@ int tegra_drm_submit(struct tegra_drm_context
>> *context,
>>         while (num_cmdbufs) {
>>                 struct drm_tegra_cmdbuf cmdbuf;
>>                 struct host1x_bo *bo;
>> +               struct tegra_bo *obj;
>> +               u64 offset;
>>
>>                 if (copy_from_user(&cmdbuf, cmdbufs, sizeof(cmdbuf))) {
>>                         err = -EFAULT;
>> @@ -374,6 +376,14 @@ int tegra_drm_submit(struct tegra_drm_context
>> *context,
>>                         goto fail;
>>                 }
>>
>> +               offset = (u64)cmdbuf.offset + (u64)cmdbuf.words *
>> sizeof(u32);
>> +               obj = host1x_to_tegra_bo(bo);
>> +
>> +               if (offset & 3 || offset > obj->gem.size) {
>> +                       err = -EINVAL;
>> +                       goto fail;
>> +               }
>> +
>>                 host1x_job_add_gather(job, bo, cmdbuf.words,
>> cmdbuf.offset);
>>                 num_cmdbufs--;
>>                 cmdbufs++;
>> @@ -381,11 +391,31 @@ int tegra_drm_submit(struct tegra_drm_context
>> *context,
>>
>>         /* copy and resolve relocations from submit */
>>         while (num_relocs--) {
>> +               struct host1x_reloc *reloc;
>> +               struct tegra_bo *obj;
>> +
>>                 err =
>> host1x_reloc_copy_from_user(&job->relocarray[num_relocs],
>>                                                   &relocs[num_relocs],
>> drm,
>>                                                   file);
>>                 if (err < 0)
>>                         goto fail;
>> +
>> +               reloc = &job->relocarray[num_relocs];
>> +               obj = host1x_to_tegra_bo(reloc->cmdbuf.bo);
>> +
>> +               if (reloc->cmdbuf.offset & 3 ||
>> +                   reloc->cmdbuf.offset > obj->gem.size) {
>
>
> This could still fail if the bo's size is not divisible by 4, even with >=
> comparison (we would overwrite the buffer by 1 to 3 bytes). I would do the
> same as in the gather case, i.e. find out the address immediately after the
> write and compare using >. Perhaps add a helper function if it makes sense.
> I also don't think the "& 3" checks are needed.

The bo-size is always a multiple of PAGE_SIZE, due to the rounding in
tegra_bo_alloc_object(), so I don't think this actually can fail. But
maybe we want to future-proof this code for a potential future where
this is not the case?

Re: [PATCH] drm/tegra: Check offsets of a submitted command buffer and of relocations

[Mikko Perttunen]

On 16.05.2017 10:32, Erik Faye-Lund wrote:
> On Tue, May 16, 2017 at 8:56 AM, Mikko Perttunen <cyndis-/1wQRMveznE@public.gmane.org> wrote:
>> On 14.05.2017 23:47, Dmitry Osipenko wrote:
>>>
>>> If commands buffer claims a number of words that is higher than its BO can
>>> fit, a kernel OOPS will be fired on the out-of-bounds BO access. This was
>>> triggered by an opentegra Xorg driver that erroneously pushed too many
>>> commands to the pushbuf. The CMDA commands buffer address is 4 bytes
>>> aligned, so check the alignment as well.
>>>
>>> Add a sanity check for the relocations in a same way.
>>>
>>> [   46.829393] Unable to handle kernel paging request at virtual address
>>> f09b2000
>>> ...
>>> [<c04a3ba4>] (host1x_job_pin) from [<c04dfcd0>]
>>> (tegra_drm_submit+0x474/0x510)
>>> [<c04dfcd0>] (tegra_drm_submit) from [<c04deea0>] (tegra_submit+0x50/0x6c)
>>> [<c04deea0>] (tegra_submit) from [<c04c07c0>] (drm_ioctl+0x1e4/0x3ec)
>>> [<c04c07c0>] (drm_ioctl) from [<c02541a0>] (do_vfs_ioctl+0x9c/0x8e4)
>>> [<c02541a0>] (do_vfs_ioctl) from [<c0254a1c>] (SyS_ioctl+0x34/0x5c)
>>> [<c0254a1c>] (SyS_ioctl) from [<c0107640>] (ret_fast_syscall+0x0/0x3c)
>>>
>>> Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>>> Reviewed-by: Erik Faye-Lund <kusmabite-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>>> ---
>>>  drivers/gpu/drm/tegra/drm.c | 30 ++++++++++++++++++++++++++++++
>>>  drivers/gpu/drm/tegra/gem.c |  5 -----
>>>  drivers/gpu/drm/tegra/gem.h |  5 +++++
>>>  3 files changed, 35 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/tegra/drm.c b/drivers/gpu/drm/tegra/drm.c
>>> index 768750226452..c5844a065681 100644
>>> --- a/drivers/gpu/drm/tegra/drm.c
>>> +++ b/drivers/gpu/drm/tegra/drm.c
>>> @@ -362,6 +362,8 @@ int tegra_drm_submit(struct tegra_drm_context
>>> *context,
>>>         while (num_cmdbufs) {
>>>                 struct drm_tegra_cmdbuf cmdbuf;
>>>                 struct host1x_bo *bo;
>>> +               struct tegra_bo *obj;
>>> +               u64 offset;
>>>
>>>                 if (copy_from_user(&cmdbuf, cmdbufs, sizeof(cmdbuf))) {
>>>                         err = -EFAULT;
>>> @@ -374,6 +376,14 @@ int tegra_drm_submit(struct tegra_drm_context
>>> *context,
>>>                         goto fail;
>>>                 }
>>>
>>> +               offset = (u64)cmdbuf.offset + (u64)cmdbuf.words *
>>> sizeof(u32);
>>> +               obj = host1x_to_tegra_bo(bo);
>>> +
>>> +               if (offset & 3 || offset > obj->gem.size) {
>>> +                       err = -EINVAL;
>>> +                       goto fail;
>>> +               }
>>> +
>>>                 host1x_job_add_gather(job, bo, cmdbuf.words,
>>> cmdbuf.offset);
>>>                 num_cmdbufs--;
>>>                 cmdbufs++;
>>> @@ -381,11 +391,31 @@ int tegra_drm_submit(struct tegra_drm_context
>>> *context,
>>>
>>>         /* copy and resolve relocations from submit */
>>>         while (num_relocs--) {
>>> +               struct host1x_reloc *reloc;
>>> +               struct tegra_bo *obj;
>>> +
>>>                 err =
>>> host1x_reloc_copy_from_user(&job->relocarray[num_relocs],
>>>                                                   &relocs[num_relocs],
>>> drm,
>>>                                                   file);
>>>                 if (err < 0)
>>>                         goto fail;
>>> +
>>> +               reloc = &job->relocarray[num_relocs];
>>> +               obj = host1x_to_tegra_bo(reloc->cmdbuf.bo);
>>> +
>>> +               if (reloc->cmdbuf.offset & 3 ||
>>> +                   reloc->cmdbuf.offset > obj->gem.size) {
>>
>>
>> This could still fail if the bo's size is not divisible by 4, even with >=
>> comparison (we would overwrite the buffer by 1 to 3 bytes). I would do the
>> same as in the gather case, i.e. find out the address immediately after the
>> write and compare using >. Perhaps add a helper function if it makes sense.
>> I also don't think the "& 3" checks are needed.
>
> The bo-size is always a multiple of PAGE_SIZE, due to the rounding in
> tegra_bo_alloc_object(), so I don't think this actually can fail. But
> maybe we want to future-proof this code for a potential future where
> this is not the case?
>

Yeah, I think whether or not this might change in the future, the code 
would be easier to understand with the change.

Re: [PATCH v2] drm/tegra: Correct idr_alloc() minimum id

[Mikko Perttunen]

Reviewed-by: Mikko Perttunen <mperttunen-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>

On 14.05.2017 23:47, Dmitry Osipenko wrote:
> The client ID 0 is reserved by the host1x/cdma to mark the timeout timer
> work as already been scheduled and context ID is used as the clients one.
> This fixes spurious CDMA timeouts.
>
> Fixes: bdd2f9cd10eb ("drm/tegra: Don't leak kernel pointer to userspace")
> Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> ---
>
> v2: Changed the commit description, now explains the cause of CDMA timeouts.
>
>  drivers/gpu/drm/tegra/drm.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/tegra/drm.c b/drivers/gpu/drm/tegra/drm.c
> index c5844a065681..489cb32453f7 100644
> --- a/drivers/gpu/drm/tegra/drm.c
> +++ b/drivers/gpu/drm/tegra/drm.c
> @@ -548,7 +548,7 @@ static int tegra_client_open(struct tegra_drm_file *fpriv,
>  	if (err < 0)
>  		return err;
>
> -	err = idr_alloc(&fpriv->contexts, context, 0, 0, GFP_KERNEL);
> +	err = idr_alloc(&fpriv->contexts, context, 1, 0, GFP_KERNEL);
>  	if (err < 0) {
>  		client->ops->close_channel(context);
>  		return err;
>

Re: [PATCH] drm/tegra: Check offsets of a submitted command buffer and of relocations

[Dmitry Osipenko]

On 16.05.2017 09:56, Mikko Perttunen wrote:
> On 14.05.2017 23:47, Dmitry Osipenko wrote:
>> +        if (reloc->cmdbuf.offset & 3 ||
>> +            reloc->cmdbuf.offset > obj->gem.size) {
> 
> This could still fail if the bo's size is not divisible by 4, even with >=
> comparison (we would overwrite the buffer by 1 to 3 bytes). I would do the same
> as in the gather case, i.e. find out the address immediately after the write and
> compare using >. Perhaps add a helper function if it makes sense. I also don't
> think the "& 3" checks are needed.
> 
Sorry, I forgot to mention the reason of the relocations cmdbuf offset alignment
requirement in the commit description. So the reason is that during the patching
of the gather in do_relocs(), the unaligned offset would cause a partial
overwrite of the command buffer word following the relocation and only a part of
the relocation address being written to the actual address word.

Re: [PATCH] drm/tegra: Check offsets of a submitted command buffer and of relocations

[Dmitry Osipenko]

On 16.05.2017 11:10, Mikko Perttunen wrote:
> 
> 
> On 16.05.2017 10:32, Erik Faye-Lund wrote:
>> On Tue, May 16, 2017 at 8:56 AM, Mikko Perttunen <cyndis-/1wQRMveznE@public.gmane.org> wrote:
>>> On 14.05.2017 23:47, Dmitry Osipenko wrote:
>>>>
>>>> If commands buffer claims a number of words that is higher than its BO can
>>>> fit, a kernel OOPS will be fired on the out-of-bounds BO access. This was
>>>> triggered by an opentegra Xorg driver that erroneously pushed too many
>>>> commands to the pushbuf. The CMDA commands buffer address is 4 bytes
>>>> aligned, so check the alignment as well.
>>>>
>>>> Add a sanity check for the relocations in a same way.
>>>>
>>>> [   46.829393] Unable to handle kernel paging request at virtual address
>>>> f09b2000
>>>> ...
>>>> [<c04a3ba4>] (host1x_job_pin) from [<c04dfcd0>]
>>>> (tegra_drm_submit+0x474/0x510)
>>>> [<c04dfcd0>] (tegra_drm_submit) from [<c04deea0>] (tegra_submit+0x50/0x6c)
>>>> [<c04deea0>] (tegra_submit) from [<c04c07c0>] (drm_ioctl+0x1e4/0x3ec)
>>>> [<c04c07c0>] (drm_ioctl) from [<c02541a0>] (do_vfs_ioctl+0x9c/0x8e4)
>>>> [<c02541a0>] (do_vfs_ioctl) from [<c0254a1c>] (SyS_ioctl+0x34/0x5c)
>>>> [<c0254a1c>] (SyS_ioctl) from [<c0107640>] (ret_fast_syscall+0x0/0x3c)
>>>>
>>>> Signed-off-by: Dmitry Osipenko <digetx-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>>>> Reviewed-by: Erik Faye-Lund <kusmabite-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>>>> ---
>>>>  drivers/gpu/drm/tegra/drm.c | 30 ++++++++++++++++++++++++++++++
>>>>  drivers/gpu/drm/tegra/gem.c |  5 -----
>>>>  drivers/gpu/drm/tegra/gem.h |  5 +++++
>>>>  3 files changed, 35 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/tegra/drm.c b/drivers/gpu/drm/tegra/drm.c
>>>> index 768750226452..c5844a065681 100644
>>>> --- a/drivers/gpu/drm/tegra/drm.c
>>>> +++ b/drivers/gpu/drm/tegra/drm.c
>>>> @@ -362,6 +362,8 @@ int tegra_drm_submit(struct tegra_drm_context
>>>> *context,
>>>>         while (num_cmdbufs) {
>>>>                 struct drm_tegra_cmdbuf cmdbuf;
>>>>                 struct host1x_bo *bo;
>>>> +               struct tegra_bo *obj;
>>>> +               u64 offset;
>>>>
>>>>                 if (copy_from_user(&cmdbuf, cmdbufs, sizeof(cmdbuf))) {
>>>>                         err = -EFAULT;
>>>> @@ -374,6 +376,14 @@ int tegra_drm_submit(struct tegra_drm_context
>>>> *context,
>>>>                         goto fail;
>>>>                 }
>>>>
>>>> +               offset = (u64)cmdbuf.offset + (u64)cmdbuf.words *
>>>> sizeof(u32);
>>>> +               obj = host1x_to_tegra_bo(bo);
>>>> +
>>>> +               if (offset & 3 || offset > obj->gem.size) {
>>>> +                       err = -EINVAL;
>>>> +                       goto fail;
>>>> +               }
>>>> +
>>>>                 host1x_job_add_gather(job, bo, cmdbuf.words,
>>>> cmdbuf.offset);
>>>>                 num_cmdbufs--;
>>>>                 cmdbufs++;
>>>> @@ -381,11 +391,31 @@ int tegra_drm_submit(struct tegra_drm_context
>>>> *context,
>>>>
>>>>         /* copy and resolve relocations from submit */
>>>>         while (num_relocs--) {
>>>> +               struct host1x_reloc *reloc;
>>>> +               struct tegra_bo *obj;
>>>> +
>>>>                 err =
>>>> host1x_reloc_copy_from_user(&job->relocarray[num_relocs],
>>>>                                                   &relocs[num_relocs],
>>>> drm,
>>>>                                                   file);
>>>>                 if (err < 0)
>>>>                         goto fail;
>>>> +
>>>> +               reloc = &job->relocarray[num_relocs];
>>>> +               obj = host1x_to_tegra_bo(reloc->cmdbuf.bo);
>>>> +
>>>> +               if (reloc->cmdbuf.offset & 3 ||
>>>> +                   reloc->cmdbuf.offset > obj->gem.size) {
>>>
>>>
>>> This could still fail if the bo's size is not divisible by 4, even with >=
>>> comparison (we would overwrite the buffer by 1 to 3 bytes). I would do the
>>> same as in the gather case, i.e. find out the address immediately after the
>>> write and compare using >. Perhaps add a helper function if it makes sense.
>>> I also don't think the "& 3" checks are needed.
>>
>> The bo-size is always a multiple of PAGE_SIZE, due to the rounding in
>> tegra_bo_alloc_object(), so I don't think this actually can fail. But
>> maybe we want to future-proof this code for a potential future where
>> this is not the case?
>>
> 
> Yeah, I think whether or not this might change in the future, the code would be
> easier to understand with the change.

It's very unlikely to me that BO size round-upping may go away. I'll add a
comment for that alignment check to make it clear in the code.

-- 
Dmitry