[PATCH v4 1/5] staging/android: add num_fences field to struct sync_file_info

[PATCH v4 1/5] staging/android: add num_fences field to struct sync_file_info

[Gustavo Padovan]

From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>

Inform userspace how many fences are in the sync_fence_info field.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
---
 drivers/staging/android/sync.c      | 2 ++
 drivers/staging/android/uapi/sync.h | 2 ++
 2 files changed, 4 insertions(+)

diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
index 3a8f210..31aa462 100644
--- a/drivers/staging/android/sync.c
+++ b/drivers/staging/android/sync.c
@@ -525,6 +525,8 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
 	if (info->status >= 0)
 		info->status = !info->status;
 
+	info->num_fences = sync_file->num_fences;
+
 	len = sizeof(struct sync_file_info);
 
 	for (i = 0; i < sync_file->num_fences; ++i) {
diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
index a0cf357..4ffb7cc 100644
--- a/drivers/staging/android/uapi/sync.h
+++ b/drivers/staging/android/uapi/sync.h
@@ -47,12 +47,14 @@ struct sync_fence_info {
  *		userspace including pt_info.
  * @name:	name of fence
  * @status:	status of fence. 1: signaled 0:active <0:error
+ * @num_fences	number of fences in the sync_file
  * @sync_fence_info: array of sync_fence_info for every fence in the sync_file
  */
 struct sync_file_info {
 	__u32	len;
 	char	name[32];
 	__s32	status;
+	__u32	num_fences;
 
 	__u8	sync_fence_info[0];
 };
-- 
2.5.0

[PATCH v4 2/5] staging/android: rename SYNC_IOC_FENCE_INFO

[Gustavo Padovan]

From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>

We don't use the 'fence' name to refer to sync_file anymore. So rename it
to SYNC_IOC_FILE_INFO.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
---
 drivers/staging/android/sync.c      | 2 +-
 drivers/staging/android/uapi/sync.h | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
index 31aa462..dc5f382 100644
--- a/drivers/staging/android/sync.c
+++ b/drivers/staging/android/sync.c
@@ -562,7 +562,7 @@ static long sync_file_ioctl(struct file *file, unsigned int cmd,
 	case SYNC_IOC_MERGE:
 		return sync_file_ioctl_merge(sync_file, arg);
 
-	case SYNC_IOC_FENCE_INFO:
+	case SYNC_IOC_FILE_INFO:
 		return sync_file_ioctl_fence_info(sync_file, arg);
 
 	default:
diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
index 4ffb7cc..dd0dd84 100644
--- a/drivers/staging/android/uapi/sync.h
+++ b/drivers/staging/android/uapi/sync.h
@@ -81,6 +81,6 @@ struct sync_file_info {
  * pt_info is a buffer containing sync_pt_infos for every sync_pt in the fence.
  * To iterate over the sync_pt_infos, use the sync_pt_info.len field.
  */
-#define SYNC_IOC_FENCE_INFO	_IOWR(SYNC_IOC_MAGIC, 2, struct sync_file_info)
+#define SYNC_IOC_FILE_INFO	_IOWR(SYNC_IOC_MAGIC, 2, struct sync_file_info)
 
 #endif /* _UAPI_LINUX_SYNC_H */
-- 
2.5.0

[PATCH v4 3/5] staging/android: remove redundant comments on sync_merge_data

[Gustavo Padovan]

From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>

struct sync_merge_data already have documentation on top of the
struct definition. No need to duplicate it.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
---
 drivers/staging/android/uapi/sync.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
index dd0dd84..f0b41ce 100644
--- a/drivers/staging/android/uapi/sync.h
+++ b/drivers/staging/android/uapi/sync.h
@@ -21,9 +21,9 @@
  * @fence:	returns the fd of the new fence to userspace
  */
 struct sync_merge_data {
-	__s32	fd2; /* fd of second fence */
-	char	name[32]; /* name of new fence */
-	__s32	fence; /* fd on newly created fence */
+	__s32	fd2;
+	char	name[32];
+	__s32	fence;
 };
 
 /**
-- 
2.5.0

[PATCH v4 4/5] staging/android: refactor SYNC_IOC_FILE_INFO

[Gustavo Padovan]

From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>

Change SYNC_IOC_FILE_INFO behaviour to avoid future API breaks and
optimize buffer allocation. In the new approach the ioctl needs to be called
twice to retrieve the array of fence_infos pointed by info->sync_fence_info.

The first call should pass num_fences = 0, the kernel will then fill
info->num_fences. Userspace receives back the number of fences and
allocates a buffer size num_fences * sizeof(struct sync_fence_info) on
info->sync_fence_info.

It then call the ioctl again passing num_fences received in info->num_fences.
The kernel checks if info->num_fences > 0 and if yes it fill
info->sync_fence_info with an array containing all fence_infos.

info->len now represents the length of the buffer sync_fence_info points
to. Also, info->sync_fence_info was converted to __u64 pointer.

An example userspace code would be:

	struct sync_file_info *info;
	int err, size, num_fences;

	info = malloc(sizeof(*info));

	memset(info, 0, sizeof(*info));

	err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
	num_fences = info->num_fences;

	if (num_fences) {
		memset(info, 0, sizeof(*info));
		size = sizeof(struct sync_fence_info) * num_fences;
		info->len = size;
		info->num_fences = num_fences;
		info->sync_fence_info = (uint64_t) calloc(num_fences,
							  sizeof(struct sync_fence_info));

		err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
	}

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
---
 drivers/staging/android/sync.c      | 56 +++++++++++++++++++++++++++++--------
 drivers/staging/android/uapi/sync.h |  9 +++---
 2 files changed, 48 insertions(+), 17 deletions(-)

diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
index dc5f382..837cff5 100644
--- a/drivers/staging/android/sync.c
+++ b/drivers/staging/android/sync.c
@@ -502,21 +502,22 @@ static int sync_fill_fence_info(struct fence *fence, void *data, int size)
 static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
 					unsigned long arg)
 {
-	struct sync_file_info *info;
+	struct sync_file_info in, *info;
+	struct sync_fence_info *fence_info;
 	__u32 size;
 	__u32 len = 0;
 	int ret, i;
 
-	if (copy_from_user(&size, (void __user *)arg, sizeof(size)))
+	if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
 		return -EFAULT;
 
-	if (size < sizeof(struct sync_file_info))
-		return -EINVAL;
+	if (in.status || strcmp(in.name, "\0"))
+		return -EFAULT;
 
-	if (size > 4096)
-		size = 4096;
+	if (in.num_fences && !in.sync_fence_info)
+		return -EFAULT;
 
-	info = kzalloc(size, GFP_KERNEL);
+	info = kzalloc(sizeof(*info), GFP_KERNEL);
 	if (!info)
 		return -ENOMEM;
 
@@ -525,24 +526,55 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
 	if (info->status >= 0)
 		info->status = !info->status;
 
-	info->num_fences = sync_file->num_fences;
+	/*
+	 * Passing num_fences = 0 means that userspace want to know how
+	 * many fences are in the sync_file to be able to allocate a buffer to
+	 * fit all sync_fence_infos and call the ioctl again with the buffer
+	 * assigned to info->sync_fence_info. The second call pass the
+	 * num_fences value received in the first call.
+	 */
+	if (!in.num_fences)
+		goto no_fences;
 
-	len = sizeof(struct sync_file_info);
+	size = sync_file->num_fences * sizeof(*fence_info);
+	if (in.len != size) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	fence_info = kzalloc(size, GFP_KERNEL);
+	if (!fence_info) {
+		ret = -ENOMEM;
+		goto out;
+	}
 
 	for (i = 0; i < sync_file->num_fences; ++i) {
 		struct fence *fence = sync_file->cbs[i].fence;
 
-		ret = sync_fill_fence_info(fence, (u8 *)info + len, size - len);
+		ret = sync_fill_fence_info(fence, (u8 *)fence_info + len,
+					   size - len);
 
-		if (ret < 0)
+		if (ret < 0) {
+			kfree(fence_info);
 			goto out;
+		}
 
 		len += ret;
 	}
 
+	if (copy_to_user((void __user *)in.sync_fence_info, fence_info, size)) {
+		ret = -EFAULT;
+		kfree(fence_info);
+		goto out;
+	}
+
 	info->len = len;
+	info->sync_fence_info = (__u64) in.sync_fence_info;
+
+no_fences:
+	info->num_fences = sync_file->num_fences;
 
-	if (copy_to_user((void __user *)arg, info, len))
+	if (copy_to_user((void __user *)arg, info, sizeof(*info)))
 		ret = -EFAULT;
 	else
 		ret = 0;
diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
index f0b41ce..9aad623 100644
--- a/drivers/staging/android/uapi/sync.h
+++ b/drivers/staging/android/uapi/sync.h
@@ -42,21 +42,20 @@ struct sync_fence_info {
 
 /**
  * struct sync_file_info - data returned from fence info ioctl
- * @len:	ioctl caller writes the size of the buffer its passing in.
- *		ioctl returns length of sync_file_info returned to
- *		userspace including pt_info.
  * @name:	name of fence
  * @status:	status of fence. 1: signaled 0:active <0:error
  * @num_fences	number of fences in the sync_file
+ * @len:	ioctl caller writes the size of the buffer its passing in.
+ *		ioctl returns length of all fence_infos summed.
  * @sync_fence_info: array of sync_fence_info for every fence in the sync_file
  */
 struct sync_file_info {
-	__u32	len;
 	char	name[32];
 	__s32	status;
 	__u32	num_fences;
+	__u32	len;
 
-	__u8	sync_fence_info[0];
+	__u64	sync_fence_info;
 };
 
 #define SYNC_IOC_MAGIC		'>'
-- 
2.5.0

[PATCH v4 5/5] staging/android: add flags member to sync ioctl structs

[Gustavo Padovan]

From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>

Play safe and add flags member to all structs. So we don't need to
break API or create new IOCTL in the future if new features that requires
flags arises.

v2: check if flags are valid (zero, in this case)

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
---
 drivers/staging/android/sync.c      | 7 ++++++-
 drivers/staging/android/uapi/sync.h | 6 ++++++
 2 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
index 837cff5..54fd5ab 100644
--- a/drivers/staging/android/sync.c
+++ b/drivers/staging/android/sync.c
@@ -445,6 +445,11 @@ static long sync_file_ioctl_merge(struct sync_file *sync_file,
 		goto err_put_fd;
 	}
 
+	if (data.flags) {
+		err = -EFAULT;
+		goto err_put_fd;
+	}
+
 	fence2 = sync_file_fdget(data.fd2);
 	if (!fence2) {
 		err = -ENOENT;
@@ -511,7 +516,7 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
 	if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
 		return -EFAULT;
 
-	if (in.status || strcmp(in.name, "\0"))
+	if (in.status || in.flags || strcmp(in.name, "\0"))
 		return -EFAULT;
 
 	if (in.num_fences && !in.sync_fence_info)
diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
index 9aad623..f56a6c2 100644
--- a/drivers/staging/android/uapi/sync.h
+++ b/drivers/staging/android/uapi/sync.h
@@ -19,11 +19,13 @@
  * @fd2:	file descriptor of second fence
  * @name:	name of new fence
  * @fence:	returns the fd of the new fence to userspace
+ * @flags:	merge_data flags
  */
 struct sync_merge_data {
 	__s32	fd2;
 	char	name[32];
 	__s32	fence;
+	__u32	flags;
 };
 
 /**
@@ -31,12 +33,14 @@ struct sync_merge_data {
  * @obj_name:		name of parent sync_timeline
  * @driver_name:	name of driver implementing the parent
  * @status:		status of the fence 0:active 1:signaled <0:error
+ * @flags:		fence_info flags
  * @timestamp_ns:	timestamp of status change in nanoseconds
  */
 struct sync_fence_info {
 	char	obj_name[32];
 	char	driver_name[32];
 	__s32	status;
+	__u32	flags;
 	__u64	timestamp_ns;
 };
 
@@ -44,6 +48,7 @@ struct sync_fence_info {
  * struct sync_file_info - data returned from fence info ioctl
  * @name:	name of fence
  * @status:	status of fence. 1: signaled 0:active <0:error
+ * @flags:	sync_file_info flags
  * @num_fences	number of fences in the sync_file
  * @len:	ioctl caller writes the size of the buffer its passing in.
  *		ioctl returns length of all fence_infos summed.
@@ -52,6 +57,7 @@ struct sync_fence_info {
 struct sync_file_info {
 	char	name[32];
 	__s32	status;
+	__u32	flags;
 	__u32	num_fences;
 	__u32	len;
 
-- 
2.5.0

[PATCH] staging/android: refactor SYNC_IOC_FILE_INFO

[Gustavo Padovan]

From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>

Change SYNC_IOC_FILE_INFO behaviour to avoid future API breaks and
optimize buffer allocation. In the new approach the ioctl needs to be called
twice to retrieve the array of fence_infos pointed by info->sync_fence_info.

The first call should pass num_fences = 0, the kernel will then fill
info->num_fences. Userspace receives back the number of fences and
allocates a buffer size num_fences * sizeof(struct sync_fence_info) on
info->sync_fence_info.

It then call the ioctl again passing num_fences received in info->num_fences.
The kernel checks if info->num_fences > 0 and if yes it fill
info->sync_fence_info with an array containing all fence_infos.

info->len now represents the length of the buffer sync_fence_info points
to. Also, info->sync_fence_info was converted to __u64 pointer.

An example userspace code would be:

	struct sync_file_info *info;
	int err, size, num_fences;

	info = malloc(sizeof(*info));

	memset(info, 0, sizeof(*info));

	err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
	num_fences = info->num_fences;

	if (num_fences) {
		memset(info, 0, sizeof(*info));
		size = sizeof(struct sync_fence_info) * num_fences;
		info->len = size;
		info->num_fences = num_fences;
		info->sync_fence_info = (uint64_t) calloc(num_fences,
							  sizeof(struct sync_fence_info));

		err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
	}

v2: fix fence_info memory leak

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
---
 drivers/staging/android/sync.c      | 52 +++++++++++++++++++++++++++++--------
 drivers/staging/android/uapi/sync.h |  9 +++----
 2 files changed, 45 insertions(+), 16 deletions(-)

diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
index dc5f382..2379f23 100644
--- a/drivers/staging/android/sync.c
+++ b/drivers/staging/android/sync.c
@@ -502,21 +502,22 @@ static int sync_fill_fence_info(struct fence *fence, void *data, int size)
 static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
 					unsigned long arg)
 {
-	struct sync_file_info *info;
+	struct sync_file_info in, *info;
+	struct sync_fence_info *fence_info = NULL;
 	__u32 size;
 	__u32 len = 0;
 	int ret, i;
 
-	if (copy_from_user(&size, (void __user *)arg, sizeof(size)))
+	if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
 		return -EFAULT;
 
-	if (size < sizeof(struct sync_file_info))
-		return -EINVAL;
+	if (in.status || strcmp(in.name, "\0"))
+		return -EFAULT;
 
-	if (size > 4096)
-		size = 4096;
+	if (in.num_fences && !in.sync_fence_info)
+		return -EFAULT;
 
-	info = kzalloc(size, GFP_KERNEL);
+	info = kzalloc(sizeof(*info), GFP_KERNEL);
 	if (!info)
 		return -ENOMEM;
 
@@ -525,14 +526,33 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
 	if (info->status >= 0)
 		info->status = !info->status;
 
-	info->num_fences = sync_file->num_fences;
+	/*
+	 * Passing num_fences = 0 means that userspace want to know how
+	 * many fences are in the sync_file to be able to allocate a buffer to
+	 * fit all sync_fence_infos and call the ioctl again with the buffer
+	 * assigned to info->sync_fence_info. The second call pass the
+	 * num_fences value received in the first call.
+	 */
+	if (!in.num_fences)
+		goto no_fences;
+
+	size = sync_file->num_fences * sizeof(*fence_info);
+	if (in.len != size) {
+		ret = -EFAULT;
+		goto out;
+	}
 
-	len = sizeof(struct sync_file_info);
+	fence_info = kzalloc(size, GFP_KERNEL);
+	if (!fence_info) {
+		ret = -ENOMEM;
+		goto out;
+	}
 
 	for (i = 0; i < sync_file->num_fences; ++i) {
 		struct fence *fence = sync_file->cbs[i].fence;
 
-		ret = sync_fill_fence_info(fence, (u8 *)info + len, size - len);
+		ret = sync_fill_fence_info(fence, (u8 *)fence_info + len,
+					   size - len);
 
 		if (ret < 0)
 			goto out;
@@ -540,14 +560,24 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
 		len += ret;
 	}
 
+	if (copy_to_user((void __user *)in.sync_fence_info, fence_info, size)) {
+		ret = -EFAULT;
+		goto out;
+	}
+
 	info->len = len;
+	info->sync_fence_info = (__u64) in.sync_fence_info;
+
+no_fences:
+	info->num_fences = sync_file->num_fences;
 
-	if (copy_to_user((void __user *)arg, info, len))
+	if (copy_to_user((void __user *)arg, info, sizeof(*info)))
 		ret = -EFAULT;
 	else
 		ret = 0;
 
 out:
+	kfree(fence_info);
 	kfree(info);
 
 	return ret;
diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
index f0b41ce..9aad623 100644
--- a/drivers/staging/android/uapi/sync.h
+++ b/drivers/staging/android/uapi/sync.h
@@ -42,21 +42,20 @@ struct sync_fence_info {
 
 /**
  * struct sync_file_info - data returned from fence info ioctl
- * @len:	ioctl caller writes the size of the buffer its passing in.
- *		ioctl returns length of sync_file_info returned to
- *		userspace including pt_info.
  * @name:	name of fence
  * @status:	status of fence. 1: signaled 0:active <0:error
  * @num_fences	number of fences in the sync_file
+ * @len:	ioctl caller writes the size of the buffer its passing in.
+ *		ioctl returns length of all fence_infos summed.
  * @sync_fence_info: array of sync_fence_info for every fence in the sync_file
  */
 struct sync_file_info {
-	__u32	len;
 	char	name[32];
 	__s32	status;
 	__u32	num_fences;
+	__u32	len;
 
-	__u8	sync_fence_info[0];
+	__u64	sync_fence_info;
 };
 
 #define SYNC_IOC_MAGIC		'>'
-- 
2.5.0

Re: [PATCH] staging/android: refactor SYNC_IOC_FILE_INFO

[Emil Velikov]

Hi Gustavo,

On 26 February 2016 at 21:00, Gustavo Padovan <gustavo@padovan.org> wrote:
> From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>
> Change SYNC_IOC_FILE_INFO behaviour to avoid future API breaks and
> optimize buffer allocation. In the new approach the ioctl needs to be called
> twice to retrieve the array of fence_infos pointed by info->sync_fence_info.
>
I might have misunderstood things but I no one says you "need" to call
it twice - you can just request a "random" amount of fences_info. Upon
return (if num_fences was non zero) it will report how many fence_info
were retrieved.

> The first call should pass num_fences = 0, the kernel will then fill
> info->num_fences. Userspace receives back the number of fences and
> allocates a buffer size num_fences * sizeof(struct sync_fence_info) on
> info->sync_fence_info.
>
> It then call the ioctl again passing num_fences received in info->num_fences.
"calls"

> The kernel checks if info->num_fences > 0 and if yes it fill
> info->sync_fence_info with an array containing all fence_infos.
>
The above sentence sounds a bit strange. I believe you meant to say
something like "Then the kernel fills the fence_infos array with data.
One should read back the actual number from info->num_fences." ?

> info->len now represents the length of the buffer sync_fence_info points
> to.
Now that I think about it, I'm wondering if there'll be a case where
len != info->num_fences * sizeof(struct sync_file_info). If that's not
possible one could just drop len and nicely simplify things.

> Also, info->sync_fence_info was converted to __u64 pointer.
>
... pointer to prevent 32bit compatibility issues.

> An example userspace code would be:
>
>         struct sync_file_info *info;
>         int err, size, num_fences;
>
>         info = malloc(sizeof(*info));
>
>         memset(info, 0, sizeof(*info));
>
>         err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
>         num_fences = info->num_fences;
>
>         if (num_fences) {
>                 memset(info, 0, sizeof(*info));
>                 size = sizeof(struct sync_fence_info) * num_fences;
>                 info->len = size;
>                 info->num_fences = num_fences;
>                 info->sync_fence_info = (uint64_t) calloc(num_fences,
>                                                           sizeof(struct sync_fence_info));
>
>                 err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
>         }
>
> v2: fix fence_info memory leak
>
> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> ---
>  drivers/staging/android/sync.c      | 52 +++++++++++++++++++++++++++++--------
>  drivers/staging/android/uapi/sync.h |  9 +++----
>  2 files changed, 45 insertions(+), 16 deletions(-)
>
> diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
> index dc5f382..2379f23 100644
> --- a/drivers/staging/android/sync.c
> +++ b/drivers/staging/android/sync.c
> @@ -502,21 +502,22 @@ static int sync_fill_fence_info(struct fence *fence, void *data, int size)
>  static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>                                         unsigned long arg)
>  {
> -       struct sync_file_info *info;
> +       struct sync_file_info in, *info;
> +       struct sync_fence_info *fence_info = NULL;
>         __u32 size;
>         __u32 len = 0;
>         int ret, i;
>
> -       if (copy_from_user(&size, (void __user *)arg, sizeof(size)))
> +       if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
s/*info/in/

>                 return -EFAULT;
>
> -       if (size < sizeof(struct sync_file_info))
> -               return -EINVAL;
> +       if (in.status || strcmp(in.name, "\0"))
Afaict these two are outputs, so we should be checking them ?

> +               return -EFAULT;
>
As originally, input validation should return -EINVAL on error.


> -       if (size > 4096)
> -               size = 4096;
> +       if (in.num_fences && !in.sync_fence_info)
> +               return -EFAULT;
>
Ditto.

> -       info = kzalloc(size, GFP_KERNEL);
> +       info = kzalloc(sizeof(*info), GFP_KERNEL);
>         if (!info)
>                 return -ENOMEM;
>
> @@ -525,14 +526,33 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>         if (info->status >= 0)
>                 info->status = !info->status;
>
> -       info->num_fences = sync_file->num_fences;
> +       /*
> +        * Passing num_fences = 0 means that userspace want to know how
> +        * many fences are in the sync_file to be able to allocate a buffer to
> +        * fit all sync_fence_infos and call the ioctl again with the buffer
> +        * assigned to info->sync_fence_info. The second call pass the
> +        * num_fences value received in the first call.
> +        */
> +       if (!in.num_fences)
> +               goto no_fences;
> +
We should clamp in.num_fences to min2(in.num_fences,
sync_file->num_fences) and use it over sync_file->num_fences though
the rest of the function. Or just bail out when the two are not the
same.

Depends on what the planned semantics are. Fwiw I'm leaning towards the former.

> +       size = sync_file->num_fences * sizeof(*fence_info);
> +       if (in.len != size) {
> +               ret = -EFAULT;
EINVAL or just drop len from the struct.

> +               goto out;
> +       }
>
> -       len = sizeof(struct sync_file_info);
> +       fence_info = kzalloc(size, GFP_KERNEL);
> +       if (!fence_info) {
> +               ret = -ENOMEM;
> +               goto out;
> +       }
>
>         for (i = 0; i < sync_file->num_fences; ++i) {
>                 struct fence *fence = sync_file->cbs[i].fence;
>
> -               ret = sync_fill_fence_info(fence, (u8 *)info + len, size - len);
> +               ret = sync_fill_fence_info(fence, (u8 *)fence_info + len,
A few comments about sync_fill_fence_info()
 - Internal function so make the second argument of the correct type -
struct sync_fence_info *
 - Drop the third argument size, as that one is always sizeof(sync_fence_info).
 - Remove the size checking in the same function and make its return type void

Then one can simplify this loop even further :-)

> +                                          size - len);
>
>                 if (ret < 0)
>                         goto out;
> @@ -540,14 +560,24 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>                 len += ret;
>         }
>
> +       if (copy_to_user((void __user *)in.sync_fence_info, fence_info, size)) {
> +               ret = -EFAULT;
> +               goto out;
> +       }
> +
>         info->len = len;
> +       info->sync_fence_info = (__u64) in.sync_fence_info;
Why the cast ?

> +
> +no_fences:
> +       info->num_fences = sync_file->num_fences;
>
> -       if (copy_to_user((void __user *)arg, info, len))
> +       if (copy_to_user((void __user *)arg, info, sizeof(*info)))
Don't know if we should be returning (copying) any other information
but info->num_fences in case of "no_fences". In case it's not clear -
I'm thinking about the data we already have in in info->name and
info->status.

Regards,
Emil

Re: [PATCH v4 5/5] staging/android: add flags member to sync ioctl structs

[Emil Velikov]

Hi Gustavo,

On 26 February 2016 at 18:31, Gustavo Padovan <gustavo@padovan.org> wrote:
> From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>
> Play safe and add flags member to all structs. So we don't need to
> break API or create new IOCTL in the future if new features that requires
> flags arises.
>
> v2: check if flags are valid (zero, in this case)
>
> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> ---
>  drivers/staging/android/sync.c      | 7 ++++++-
>  drivers/staging/android/uapi/sync.h | 6 ++++++
>  2 files changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
> index 837cff5..54fd5ab 100644
> --- a/drivers/staging/android/sync.c
> +++ b/drivers/staging/android/sync.c
> @@ -445,6 +445,11 @@ static long sync_file_ioctl_merge(struct sync_file *sync_file,
>                 goto err_put_fd;
>         }
>
> +       if (data.flags) {
> +               err = -EFAULT;
-EINVAL ?

> +               goto err_put_fd;
> +       }
> +
>         fence2 = sync_file_fdget(data.fd2);
>         if (!fence2) {
>                 err = -ENOENT;
> @@ -511,7 +516,7 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>         if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
>                 return -EFAULT;
>
> -       if (in.status || strcmp(in.name, "\0"))
> +       if (in.status || in.flags || strcmp(in.name, "\0"))
>                 return -EFAULT;
-EINVAL ?

>
>         if (in.num_fences && !in.sync_fence_info)
> diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
> index 9aad623..f56a6c2 100644
> --- a/drivers/staging/android/uapi/sync.h
> +++ b/drivers/staging/android/uapi/sync.h
> @@ -19,11 +19,13 @@
>   * @fd2:       file descriptor of second fence
>   * @name:      name of new fence
>   * @fence:     returns the fd of the new fence to userspace
> + * @flags:     merge_data flags
>   */
>  struct sync_merge_data {
>         __s32   fd2;
>         char    name[32];
>         __s32   fence;
> +       __u32   flags;
The overall size of the struct is not multiple of 64bit, so things
will end up badly if we decide to extend it in the future. Even if
there's a small chance that update will be needed, we might as well
pad it now (and check the padding for zero, returning -EINVAL).

>  };
>
>  /**
> @@ -31,12 +33,14 @@ struct sync_merge_data {
>   * @obj_name:          name of parent sync_timeline
>   * @driver_name:       name of driver implementing the parent
>   * @status:            status of the fence 0:active 1:signaled <0:error
> + * @flags:             fence_info flags
>   * @timestamp_ns:      timestamp of status change in nanoseconds
>   */
>  struct sync_fence_info {
>         char    obj_name[32];
>         char    driver_name[32];
>         __s32   status;
> +       __u32   flags;
>         __u64   timestamp_ns;
Should we be doing some form of validation in sync_fill_fence_info()
of 'flags' ?

Regards,
Emil

Re: [PATCH] staging/android: refactor SYNC_IOC_FILE_INFO

[Gustavo Padovan]

Hi Emil,

2016-02-27 Emil Velikov <emil.l.velikov@gmail.com>:

> Hi Gustavo,
> 
> On 26 February 2016 at 21:00, Gustavo Padovan <gustavo@padovan.org> wrote:
> > From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> >
> > Change SYNC_IOC_FILE_INFO behaviour to avoid future API breaks and
> > optimize buffer allocation. In the new approach the ioctl needs to be called
> > twice to retrieve the array of fence_infos pointed by info->sync_fence_info.
> >
> I might have misunderstood things but I no one says you "need" to call
> it twice - you can just request a "random" amount of fences_info. Upon
> return (if num_fences was non zero) it will report how many fence_info
> were retrieved.

Right, I don't see any problem doing it in one request, I just didn't 
think about that in the new proposal. I'll update the code and commit
message accordinly.

> 
> > The first call should pass num_fences = 0, the kernel will then fill
> > info->num_fences. Userspace receives back the number of fences and
> > allocates a buffer size num_fences * sizeof(struct sync_fence_info) on
> > info->sync_fence_info.
> >
> > It then call the ioctl again passing num_fences received in info->num_fences.
> "calls"
> 
> > The kernel checks if info->num_fences > 0 and if yes it fill
> > info->sync_fence_info with an array containing all fence_infos.
> >
> The above sentence sounds a bit strange. I believe you meant to say
> something like "Then the kernel fills the fence_infos array with data.
> One should read back the actual number from info->num_fences." ?
> 
> > info->len now represents the length of the buffer sync_fence_info points
> > to.
> Now that I think about it, I'm wondering if there'll be a case where
> len != info->num_fences * sizeof(struct sync_file_info). If that's not
> possible one could just drop len and nicely simplify things.
> 
> > Also, info->sync_fence_info was converted to __u64 pointer.
> >
> ... pointer to prevent 32bit compatibility issues.
> 
> > An example userspace code would be:
> >
> >         struct sync_file_info *info;
> >         int err, size, num_fences;
> >
> >         info = malloc(sizeof(*info));
> >
> >         memset(info, 0, sizeof(*info));
> >
> >         err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
> >         num_fences = info->num_fences;
> >
> >         if (num_fences) {
> >                 memset(info, 0, sizeof(*info));
> >                 size = sizeof(struct sync_fence_info) * num_fences;
> >                 info->len = size;
> >                 info->num_fences = num_fences;
> >                 info->sync_fence_info = (uint64_t) calloc(num_fences,
> >                                                           sizeof(struct sync_fence_info));
> >
> >                 err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
> >         }
> >
> > v2: fix fence_info memory leak
> >
> > Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> > ---
> >  drivers/staging/android/sync.c      | 52 +++++++++++++++++++++++++++++--------
> >  drivers/staging/android/uapi/sync.h |  9 +++----
> >  2 files changed, 45 insertions(+), 16 deletions(-)
> >
> > diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
> > index dc5f382..2379f23 100644
> > --- a/drivers/staging/android/sync.c
> > +++ b/drivers/staging/android/sync.c
> > @@ -502,21 +502,22 @@ static int sync_fill_fence_info(struct fence *fence, void *data, int size)
> >  static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >                                         unsigned long arg)
> >  {
> > -       struct sync_file_info *info;
> > +       struct sync_file_info in, *info;
> > +       struct sync_fence_info *fence_info = NULL;
> >         __u32 size;
> >         __u32 len = 0;
> >         int ret, i;
> >
> > -       if (copy_from_user(&size, (void __user *)arg, sizeof(size)))
> > +       if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
> s/*info/in/
> 
> >                 return -EFAULT;
> >
> > -       if (size < sizeof(struct sync_file_info))
> > -               return -EINVAL;
> > +       if (in.status || strcmp(in.name, "\0"))
> Afaict these two are outputs, so we should be checking them ?

Hmm. Maybe not.

> 
> > +               return -EFAULT;
> >
> As originally, input validation should return -EINVAL on error.
> 
> 
> > -       if (size > 4096)
> > -               size = 4096;
> > +       if (in.num_fences && !in.sync_fence_info)
> > +               return -EFAULT;
> >
> Ditto.
> 
> > -       info = kzalloc(size, GFP_KERNEL);
> > +       info = kzalloc(sizeof(*info), GFP_KERNEL);
> >         if (!info)
> >                 return -ENOMEM;
> >
> > @@ -525,14 +526,33 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >         if (info->status >= 0)
> >                 info->status = !info->status;
> >
> > -       info->num_fences = sync_file->num_fences;
> > +       /*
> > +        * Passing num_fences = 0 means that userspace want to know how
> > +        * many fences are in the sync_file to be able to allocate a buffer to
> > +        * fit all sync_fence_infos and call the ioctl again with the buffer
> > +        * assigned to info->sync_fence_info. The second call pass the
> > +        * num_fences value received in the first call.
> > +        */
> > +       if (!in.num_fences)
> > +               goto no_fences;
> > +
> We should clamp in.num_fences to min2(in.num_fences,
> sync_file->num_fences) and use it over sync_file->num_fences though
> the rest of the function. Or just bail out when the two are not the
> same.
> 
> Depends on what the planned semantics are. Fwiw I'm leaning towards the former.

If num_fences received is smaller than the actual num_fences I think we
should fails, otherwise we should just fill the buffer with all
fence_infos...

> 
> > +       size = sync_file->num_fences * sizeof(*fence_info);
> > +       if (in.len != size) {
> > +               ret = -EFAULT;
> EINVAL or just drop len from the struct.

...so this check now would be in.len < size.

> 
> > +               goto out;
> > +       }
> >
> > -       len = sizeof(struct sync_file_info);
> > +       fence_info = kzalloc(size, GFP_KERNEL);
> > +       if (!fence_info) {
> > +               ret = -ENOMEM;
> > +               goto out;
> > +       }
> >
> >         for (i = 0; i < sync_file->num_fences; ++i) {
> >                 struct fence *fence = sync_file->cbs[i].fence;
> >
> > -               ret = sync_fill_fence_info(fence, (u8 *)info + len, size - len);
> > +               ret = sync_fill_fence_info(fence, (u8 *)fence_info + len,
> A few comments about sync_fill_fence_info()
>  - Internal function so make the second argument of the correct type -
> struct sync_fence_info *
>  - Drop the third argument size, as that one is always sizeof(sync_fence_info).
>  - Remove the size checking in the same function and make its return type void
> 
> Then one can simplify this loop even further :-)

Sounds good to me.

> 
> > +                                          size - len);
> >
> >                 if (ret < 0)
> >                         goto out;
> > @@ -540,14 +560,24 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >                 len += ret;
> >         }
> >
> > +       if (copy_to_user((void __user *)in.sync_fence_info, fence_info, size)) {
> > +               ret = -EFAULT;
> > +               goto out;
> > +       }
> > +
> >         info->len = len;
> > +       info->sync_fence_info = (__u64) in.sync_fence_info;
> Why the cast ?
> 
> > +
> > +no_fences:
> > +       info->num_fences = sync_file->num_fences;
> >
> > -       if (copy_to_user((void __user *)arg, info, len))
> > +       if (copy_to_user((void __user *)arg, info, sizeof(*info)))
> Don't know if we should be returning (copying) any other information
> but info->num_fences in case of "no_fences". In case it's not clear -
> I'm thinking about the data we already have in in info->name and
> info->status.

Userspace might want to know all info about the sync_file but
sync_fence_info.

	Gustavo

Re: [PATCH v4 5/5] staging/android: add flags member to sync ioctl structs

[Gustavo Padovan]

Hi Emil,

2016-02-27 Emil Velikov <emil.l.velikov@gmail.com>:

> Hi Gustavo,
> 
> On 26 February 2016 at 18:31, Gustavo Padovan <gustavo@padovan.org> wrote:
> > From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> >
> > Play safe and add flags member to all structs. So we don't need to
> > break API or create new IOCTL in the future if new features that requires
> > flags arises.
> >
> > v2: check if flags are valid (zero, in this case)
> >
> > Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> > ---
> >  drivers/staging/android/sync.c      | 7 ++++++-
> >  drivers/staging/android/uapi/sync.h | 6 ++++++
> >  2 files changed, 12 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
> > index 837cff5..54fd5ab 100644
> > --- a/drivers/staging/android/sync.c
> > +++ b/drivers/staging/android/sync.c
> > @@ -445,6 +445,11 @@ static long sync_file_ioctl_merge(struct sync_file *sync_file,
> >                 goto err_put_fd;
> >         }
> >
> > +       if (data.flags) {
> > +               err = -EFAULT;
> -EINVAL ?
> 
> > +               goto err_put_fd;
> > +       }
> > +
> >         fence2 = sync_file_fdget(data.fd2);
> >         if (!fence2) {
> >                 err = -ENOENT;
> > @@ -511,7 +516,7 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >         if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
> >                 return -EFAULT;
> >
> > -       if (in.status || strcmp(in.name, "\0"))
> > +       if (in.status || in.flags || strcmp(in.name, "\0"))
> >                 return -EFAULT;
> -EINVAL ?
> 
> >
> >         if (in.num_fences && !in.sync_fence_info)
> > diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
> > index 9aad623..f56a6c2 100644
> > --- a/drivers/staging/android/uapi/sync.h
> > +++ b/drivers/staging/android/uapi/sync.h
> > @@ -19,11 +19,13 @@
> >   * @fd2:       file descriptor of second fence
> >   * @name:      name of new fence
> >   * @fence:     returns the fd of the new fence to userspace
> > + * @flags:     merge_data flags
> >   */
> >  struct sync_merge_data {
> >         __s32   fd2;
> >         char    name[32];
> >         __s32   fence;
> > +       __u32   flags;
> The overall size of the struct is not multiple of 64bit, so things
> will end up badly if we decide to extend it in the future. Even if
> there's a small chance that update will be needed, we might as well
> pad it now (and check the padding for zero, returning -EINVAL).

I think name could be the first field here.

> 
> >  };
> >
> >  /**
> > @@ -31,12 +33,14 @@ struct sync_merge_data {
> >   * @obj_name:          name of parent sync_timeline
> >   * @driver_name:       name of driver implementing the parent
> >   * @status:            status of the fence 0:active 1:signaled <0:error
> > + * @flags:             fence_info flags
> >   * @timestamp_ns:      timestamp of status change in nanoseconds
> >   */
> >  struct sync_fence_info {
> >         char    obj_name[32];
> >         char    driver_name[32];
> >         __s32   status;
> > +       __u32   flags;
> >         __u64   timestamp_ns;
> Should we be doing some form of validation in sync_fill_fence_info()
> of 'flags' ?

Do you think it is necessary? The kernel allocates a zero'ed buffer to
fill sync_fence_info array.

	Gustavo

Re: [PATCH v4 5/5] staging/android: add flags member to sync ioctl structs

[Emil Velikov]

On 27 February 2016 at 15:27, Gustavo Padovan
<gustavo.padovan@collabora.co.uk> wrote:
> Hi Emil,
>
> 2016-02-27 Emil Velikov <emil.l.velikov@gmail.com>:
>
>> Hi Gustavo,
>>
>> On 26 February 2016 at 18:31, Gustavo Padovan <gustavo@padovan.org> wrote:
>> > From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>> >
>> > Play safe and add flags member to all structs. So we don't need to
>> > break API or create new IOCTL in the future if new features that requires
>> > flags arises.
>> >
>> > v2: check if flags are valid (zero, in this case)
>> >
>> > Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>> > ---
>> >  drivers/staging/android/sync.c      | 7 ++++++-
>> >  drivers/staging/android/uapi/sync.h | 6 ++++++
>> >  2 files changed, 12 insertions(+), 1 deletion(-)
>> >
>> > diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
>> > index 837cff5..54fd5ab 100644
>> > --- a/drivers/staging/android/sync.c
>> > +++ b/drivers/staging/android/sync.c
>> > @@ -445,6 +445,11 @@ static long sync_file_ioctl_merge(struct sync_file *sync_file,
>> >                 goto err_put_fd;
>> >         }
>> >
>> > +       if (data.flags) {
>> > +               err = -EFAULT;
>> -EINVAL ?
>>
>> > +               goto err_put_fd;
>> > +       }
>> > +
>> >         fence2 = sync_file_fdget(data.fd2);
>> >         if (!fence2) {
>> >                 err = -ENOENT;
>> > @@ -511,7 +516,7 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>> >         if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
>> >                 return -EFAULT;
>> >
>> > -       if (in.status || strcmp(in.name, "\0"))
>> > +       if (in.status || in.flags || strcmp(in.name, "\0"))
>> >                 return -EFAULT;
>> -EINVAL ?
>>
>> >
>> >         if (in.num_fences && !in.sync_fence_info)
>> > diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
>> > index 9aad623..f56a6c2 100644
>> > --- a/drivers/staging/android/uapi/sync.h
>> > +++ b/drivers/staging/android/uapi/sync.h
>> > @@ -19,11 +19,13 @@
>> >   * @fd2:       file descriptor of second fence
>> >   * @name:      name of new fence
>> >   * @fence:     returns the fd of the new fence to userspace
>> > + * @flags:     merge_data flags
>> >   */
>> >  struct sync_merge_data {
>> >         __s32   fd2;
>> >         char    name[32];
>> >         __s32   fence;
>> > +       __u32   flags;
>> The overall size of the struct is not multiple of 64bit, so things
>> will end up badly if we decide to extend it in the future. Even if
>> there's a small chance that update will be needed, we might as well
>> pad it now (and check the padding for zero, returning -EINVAL).
>
> I think name could be the first field here.
>
Up-to you really. I'm afraid that it doesn't resolve the issue :-(
As a test add a u64 value at the end of the struct and check the
output of pahole for 32 and 64 bit build.

>>
>> >  };
>> >
>> >  /**
>> > @@ -31,12 +33,14 @@ struct sync_merge_data {
>> >   * @obj_name:          name of parent sync_timeline
>> >   * @driver_name:       name of driver implementing the parent
>> >   * @status:            status of the fence 0:active 1:signaled <0:error
>> > + * @flags:             fence_info flags
>> >   * @timestamp_ns:      timestamp of status change in nanoseconds
>> >   */
>> >  struct sync_fence_info {
>> >         char    obj_name[32];
>> >         char    driver_name[32];
>> >         __s32   status;
>> > +       __u32   flags;
>> >         __u64   timestamp_ns;
>> Should we be doing some form of validation in sync_fill_fence_info()
>> of 'flags' ?
>
> Do you think it is necessary? The kernel allocates a zero'ed buffer to
> fill sync_fence_info array.
>
Good point. Missed out the z in kzalloc :-)

-Emil

Re: [PATCH] staging/android: refactor SYNC_IOC_FILE_INFO

[Maarten Lankhorst]

Op 26-02-16 om 22:00 schreef Gustavo Padovan:
> From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>
> Change SYNC_IOC_FILE_INFO behaviour to avoid future API breaks and
> optimize buffer allocation. In the new approach the ioctl needs to be called
> twice to retrieve the array of fence_infos pointed by info->sync_fence_info.
>
> The first call should pass num_fences = 0, the kernel will then fill
> info->num_fences. Userspace receives back the number of fences and
> allocates a buffer size num_fences * sizeof(struct sync_fence_info) on
> info->sync_fence_info.
>
> It then call the ioctl again passing num_fences received in info->num_fences.
> The kernel checks if info->num_fences > 0 and if yes it fill
> info->sync_fence_info with an array containing all fence_infos.
>
> info->len now represents the length of the buffer sync_fence_info points
> to. Also, info->sync_fence_info was converted to __u64 pointer.
>
> An example userspace code would be:
>
> 	struct sync_file_info *info;
> 	int err, size, num_fences;
>
> 	info = malloc(sizeof(*info));
>
> 	memset(info, 0, sizeof(*info));
>
> 	err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
> 	num_fences = info->num_fences;
>
> 	if (num_fences) {
> 		memset(info, 0, sizeof(*info));
Would this memset still be needed if we didn't check for nulls in info->status and info->name ?

Seems to me that it could be skipped in that case.
> 		size = sizeof(struct sync_fence_info) * num_fences;
> 		info->len = size;
> 		info->num_fences = num_fences;
> 		info->sync_fence_info = (uint64_t) calloc(num_fences,
> 							  sizeof(struct sync_fence_info));
>
> 		err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
> 	}
>
> v2: fix fence_info memory leak
>
> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> ---
>  drivers/staging/android/sync.c      | 52 +++++++++++++++++++++++++++++--------
>  drivers/staging/android/uapi/sync.h |  9 +++----
>  2 files changed, 45 insertions(+), 16 deletions(-)
>
> diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
> index dc5f382..2379f23 100644
> --- a/drivers/staging/android/sync.c
> +++ b/drivers/staging/android/sync.c
> @@ -502,21 +502,22 @@ static int sync_fill_fence_info(struct fence *fence, void *data, int size)
>  static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>  					unsigned long arg)
>  {
> -	struct sync_file_info *info;
> +	struct sync_file_info in, *info;
> +	struct sync_fence_info *fence_info = NULL;
>  	__u32 size;
>  	__u32 len = 0;
= 0 unneeded.
>  	int ret, i;
>  
> -	if (copy_from_user(&size, (void __user *)arg, sizeof(size)))
> +	if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
>  		return -EFAULT;
>  
> -	if (size < sizeof(struct sync_file_info))
> -		return -EINVAL;
> +	if (in.status || strcmp(in.name, "\0"))
> +		return -EFAULT;
These members always get written by the fence ioctl, I'm not sure it adds value to have them explicitly zero'd out by userspace.
> -	if (size > 4096)
> -		size = 4096;
> +	if (in.num_fences && !in.sync_fence_info)
> +		return -EFAULT;
This check is unneeded, it will happen in the copy_to_user call anyway.
> -	info = kzalloc(size, GFP_KERNEL);
> +	info = kzalloc(sizeof(*info), GFP_KERNEL);
>  	if (!info)
>  		return -ENOMEM;
>  
> @@ -525,14 +526,33 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>  	if (info->status >= 0)
>  		info->status = !info->status;
>  
> -	info->num_fences = sync_file->num_fences;
> +	/*
> +	 * Passing num_fences = 0 means that userspace want to know how
> +	 * many fences are in the sync_file to be able to allocate a buffer to
> +	 * fit all sync_fence_infos and call the ioctl again with the buffer
> +	 * assigned to info->sync_fence_info. The second call pass the
> +	 * num_fences value received in the first call.
> +	 */
> +	if (!in.num_fences)
> +		goto no_fences;
> +
> +	size = sync_file->num_fences * sizeof(*fence_info);
> +	if (in.len != size) {
> +		ret = -EFAULT;
> +		goto out;
> +	}
Maybe check for in.len < size, and set set to size?


> -	len = sizeof(struct sync_file_info);
> +	fence_info = kzalloc(size, GFP_KERNEL);
> +	if (!fence_info) {
> +		ret = -ENOMEM;
> +		goto out;
> +	}
>  
>  	for (i = 0; i < sync_file->num_fences; ++i) {
>  		struct fence *fence = sync_file->cbs[i].fence;
>  
> -		ret = sync_fill_fence_info(fence, (u8 *)info + len, size - len);
> +		ret = sync_fill_fence_info(fence, (u8 *)fence_info + len,
> +					   size - len);
>  
>  		if (ret < 0)
>  			goto out;
> @@ -540,14 +560,24 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>  		len += ret;
>  	}
>  
> +	if (copy_to_user((void __user *)in.sync_fence_info, fence_info, size)) {
> +		ret = -EFAULT;
> +		goto out;
> +	}
> +
>  	info->len = len;
> +	info->sync_fence_info = (__u64) in.sync_fence_info;
> +
> +no_fences:
> +	info->num_fences = sync_file->num_fences;
>  
> -	if (copy_to_user((void __user *)arg, info, len))
> +	if (copy_to_user((void __user *)arg, info, sizeof(*info)))
>  		ret = -EFAULT;
>  	else
>  		ret = 0;
>  
>  out:
> +	kfree(fence_info);
>  	kfree(info);
>  
>  	return ret;
> diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
> index f0b41ce..9aad623 100644
> --- a/drivers/staging/android/uapi/sync.h
> +++ b/drivers/staging/android/uapi/sync.h
> @@ -42,21 +42,20 @@ struct sync_fence_info {
>  
>  /**
>   * struct sync_file_info - data returned from fence info ioctl
> - * @len:	ioctl caller writes the size of the buffer its passing in.
> - *		ioctl returns length of sync_file_info returned to
> - *		userspace including pt_info.
>   * @name:	name of fence
>   * @status:	status of fence. 1: signaled 0:active <0:error
>   * @num_fences	number of fences in the sync_file
> + * @len:	ioctl caller writes the size of the buffer its passing in.
> + *		ioctl returns length of all fence_infos summed.
>   * @sync_fence_info: array of sync_fence_info for every fence in the sync_file
>   */
>  struct sync_file_info {
> -	__u32	len;
>  	char	name[32];
>  	__s32	status;
>  	__u32	num_fences;
> +	__u32	len;
>  
> -	__u8	sync_fence_info[0];
> +	__u64	sync_fence_info;
>  };
>  
>  #define SYNC_IOC_MAGIC		'>'
Not sure if len adds anything here, userspace knows to allocate num_fences * sizeof(struct sync_fence_info);

It could probably be removed since num_fences would do the same.

~Maarten

Re: [PATCH v4 5/5] staging/android: add flags member to sync ioctl structs

[Maarten Lankhorst]

Op 26-02-16 om 19:31 schreef Gustavo Padovan:
> From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>
> Play safe and add flags member to all structs. So we don't need to
> break API or create new IOCTL in the future if new features that requires
> flags arises.
>
> v2: check if flags are valid (zero, in this case)
>
> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> ---
>  drivers/staging/android/sync.c      | 7 ++++++-
>  drivers/staging/android/uapi/sync.h | 6 ++++++
>  2 files changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
> index 837cff5..54fd5ab 100644
> --- a/drivers/staging/android/sync.c
> +++ b/drivers/staging/android/sync.c
> @@ -445,6 +445,11 @@ static long sync_file_ioctl_merge(struct sync_file *sync_file,
>  		goto err_put_fd;
>  	}
>  
> +	if (data.flags) {
> +		err = -EFAULT;
> +		goto err_put_fd;
> +	}
-EINVAL
>  	fence2 = sync_file_fdget(data.fd2);
>  	if (!fence2) {
>  		err = -ENOENT;
> @@ -511,7 +516,7 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>  	if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
>  		return -EFAULT;
>  
> -	if (in.status || strcmp(in.name, "\0"))
> +	if (in.status || in.flags || strcmp(in.name, "\0"))
>  		return -EFAULT;
-EINVAL
>  	if (in.num_fences && !in.sync_fence_info)
> diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
> index 9aad623..f56a6c2 100644
> --- a/drivers/staging/android/uapi/sync.h
> +++ b/drivers/staging/android/uapi/sync.h
> @@ -19,11 +19,13 @@
>   * @fd2:	file descriptor of second fence
>   * @name:	name of new fence
>   * @fence:	returns the fd of the new fence to userspace
> + * @flags:	merge_data flags
>   */
>  struct sync_merge_data {
>  	__s32	fd2;
>  	char	name[32];
>  	__s32	fence;
> +	__u32	flags;
>  };
>  
>  /**
> @@ -31,12 +33,14 @@ struct sync_merge_data {
>   * @obj_name:		name of parent sync_timeline
>   * @driver_name:	name of driver implementing the parent
>   * @status:		status of the fence 0:active 1:signaled <0:error
> + * @flags:		fence_info flags
>   * @timestamp_ns:	timestamp of status change in nanoseconds
>   */
>  struct sync_fence_info {
>  	char	obj_name[32];
>  	char	driver_name[32];
>  	__s32	status;
> +	__u32	flags;
>  	__u64	timestamp_ns;
>  };
>  
> @@ -44,6 +48,7 @@ struct sync_fence_info {
>   * struct sync_file_info - data returned from fence info ioctl
>   * @name:	name of fence
>   * @status:	status of fence. 1: signaled 0:active <0:error
> + * @flags:	sync_file_info flags
>   * @num_fences	number of fences in the sync_file
>   * @len:	ioctl caller writes the size of the buffer its passing in.
>   *		ioctl returns length of all fence_infos summed.
> @@ -52,6 +57,7 @@ struct sync_fence_info {
>  struct sync_file_info {
>  	char	name[32];
>  	__s32	status;
> +	__u32	flags;
>  	__u32	num_fences;
>  	__u32	len;
>  

Re: [PATCH v4 3/5] staging/android: remove redundant comments on sync_merge_data

[Maarten Lankhorst]

Op 26-02-16 om 19:31 schreef Gustavo Padovan:
> From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>
> struct sync_merge_data already have documentation on top of the
> struct definition. No need to duplicate it.
>
> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> ---
>  drivers/staging/android/uapi/sync.h | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
> index dd0dd84..f0b41ce 100644
> --- a/drivers/staging/android/uapi/sync.h
> +++ b/drivers/staging/android/uapi/sync.h
> @@ -21,9 +21,9 @@
>   * @fence:	returns the fd of the new fence to userspace
>   */
>  struct sync_merge_data {
> -	__s32	fd2; /* fd of second fence */
> -	char	name[32]; /* name of new fence */
> -	__s32	fence; /* fd on newly created fence */
> +	__s32	fd2;
> +	char	name[32];
> +	__s32	fence;
>  };
>  
>  /**
For the first 3 patches:

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

Re: [PATCH] staging/android: refactor SYNC_IOC_FILE_INFO

[Emil Velikov]

Hi Gustavo,

On 27 February 2016 at 15:25, Gustavo Padovan
<gustavo.padovan@collabora.co.uk> wrote:
> Hi Emil,
>
> 2016-02-27 Emil Velikov <emil.l.velikov@gmail.com>:
>
>> Hi Gustavo,
>>
>> On 26 February 2016 at 21:00, Gustavo Padovan <gustavo@padovan.org> wrote:
>> > From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>> >
>> > Change SYNC_IOC_FILE_INFO behaviour to avoid future API breaks and
>> > optimize buffer allocation. In the new approach the ioctl needs to be called
>> > twice to retrieve the array of fence_infos pointed by info->sync_fence_info.
>> >
>> I might have misunderstood things but I no one says you "need" to call
>> it twice - you can just request a "random" amount of fences_info. Upon
>> return (if num_fences was non zero) it will report how many fence_info
>> were retrieved.
>
> Right, I don't see any problem doing it in one request, I just didn't
> think about that in the new proposal. I'll update the code and commit
> message accordinly.
>
>>
>> > The first call should pass num_fences = 0, the kernel will then fill
>> > info->num_fences. Userspace receives back the number of fences and
>> > allocates a buffer size num_fences * sizeof(struct sync_fence_info) on
>> > info->sync_fence_info.
>> >
>> > It then call the ioctl again passing num_fences received in info->num_fences.
>> "calls"
>>
>> > The kernel checks if info->num_fences > 0 and if yes it fill
>> > info->sync_fence_info with an array containing all fence_infos.
>> >
>> The above sentence sounds a bit strange. I believe you meant to say
>> something like "Then the kernel fills the fence_infos array with data.
>> One should read back the actual number from info->num_fences." ?
>>
>> > info->len now represents the length of the buffer sync_fence_info points
>> > to.
>> Now that I think about it, I'm wondering if there'll be a case where
>> len != info->num_fences * sizeof(struct sync_file_info). If that's not
>> possible one could just drop len and nicely simplify things.
>>
>> > Also, info->sync_fence_info was converted to __u64 pointer.
>> >
>> ... pointer to prevent 32bit compatibility issues.
>>
>> > An example userspace code would be:
>> >
>> >         struct sync_file_info *info;
>> >         int err, size, num_fences;
>> >
>> >         info = malloc(sizeof(*info));
>> >
>> >         memset(info, 0, sizeof(*info));
>> >
>> >         err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
>> >         num_fences = info->num_fences;
>> >
>> >         if (num_fences) {
>> >                 memset(info, 0, sizeof(*info));
>> >                 size = sizeof(struct sync_fence_info) * num_fences;
>> >                 info->len = size;
>> >                 info->num_fences = num_fences;
>> >                 info->sync_fence_info = (uint64_t) calloc(num_fences,
>> >                                                           sizeof(struct sync_fence_info));
>> >
>> >                 err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
>> >         }
>> >
>> > v2: fix fence_info memory leak
>> >
>> > Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>> > ---
>> >  drivers/staging/android/sync.c      | 52 +++++++++++++++++++++++++++++--------
>> >  drivers/staging/android/uapi/sync.h |  9 +++----
>> >  2 files changed, 45 insertions(+), 16 deletions(-)
>> >
>> > diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
>> > index dc5f382..2379f23 100644
>> > --- a/drivers/staging/android/sync.c
>> > +++ b/drivers/staging/android/sync.c
>> > @@ -502,21 +502,22 @@ static int sync_fill_fence_info(struct fence *fence, void *data, int size)
>> >  static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>> >                                         unsigned long arg)
>> >  {
>> > -       struct sync_file_info *info;
>> > +       struct sync_file_info in, *info;
>> > +       struct sync_fence_info *fence_info = NULL;
>> >         __u32 size;
>> >         __u32 len = 0;
>> >         int ret, i;
>> >
>> > -       if (copy_from_user(&size, (void __user *)arg, sizeof(size)))
>> > +       if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
>> s/*info/in/
>>
>> >                 return -EFAULT;
>> >
>> > -       if (size < sizeof(struct sync_file_info))
>> > -               return -EINVAL;
>> > +       if (in.status || strcmp(in.name, "\0"))
>> Afaict these two are outputs, so we should be checking them ?
>
> Hmm. Maybe not.
>
>>
>> > +               return -EFAULT;
>> >
>> As originally, input validation should return -EINVAL on error.
>>
>>
>> > -       if (size > 4096)
>> > -               size = 4096;
>> > +       if (in.num_fences && !in.sync_fence_info)
>> > +               return -EFAULT;
>> >
>> Ditto.
>>
>> > -       info = kzalloc(size, GFP_KERNEL);
>> > +       info = kzalloc(sizeof(*info), GFP_KERNEL);
>> >         if (!info)
>> >                 return -ENOMEM;
>> >
>> > @@ -525,14 +526,33 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>> >         if (info->status >= 0)
>> >                 info->status = !info->status;
>> >
>> > -       info->num_fences = sync_file->num_fences;
>> > +       /*
>> > +        * Passing num_fences = 0 means that userspace want to know how
>> > +        * many fences are in the sync_file to be able to allocate a buffer to
>> > +        * fit all sync_fence_infos and call the ioctl again with the buffer
>> > +        * assigned to info->sync_fence_info. The second call pass the
>> > +        * num_fences value received in the first call.
>> > +        */
>> > +       if (!in.num_fences)
>> > +               goto no_fences;
>> > +
>> We should clamp in.num_fences to min2(in.num_fences,
>> sync_file->num_fences) and use it over sync_file->num_fences though
>> the rest of the function. Or just bail out when the two are not the
>> same.
>>
>> Depends on what the planned semantics are. Fwiw I'm leaning towards the former.
>
> If num_fences received is smaller than the actual num_fences I think we
> should fails, otherwise we should just fill the buffer with all
> fence_infos...
>
Fair enough. Just make sure that this is clearly explained to the use
- manpages, other ?

>>
>> > +       size = sync_file->num_fences * sizeof(*fence_info);
>> > +       if (in.len != size) {
>> > +               ret = -EFAULT;
>> EINVAL or just drop len from the struct.
>
> ...so this check now would be in.len < size.
>
>>
>> > +               goto out;
>> > +       }
>> >
>> > -       len = sizeof(struct sync_file_info);
>> > +       fence_info = kzalloc(size, GFP_KERNEL);
>> > +       if (!fence_info) {
>> > +               ret = -ENOMEM;
>> > +               goto out;
>> > +       }
>> >
>> >         for (i = 0; i < sync_file->num_fences; ++i) {
>> >                 struct fence *fence = sync_file->cbs[i].fence;
>> >
>> > -               ret = sync_fill_fence_info(fence, (u8 *)info + len, size - len);
>> > +               ret = sync_fill_fence_info(fence, (u8 *)fence_info + len,
>> A few comments about sync_fill_fence_info()
>>  - Internal function so make the second argument of the correct type -
>> struct sync_fence_info *
>>  - Drop the third argument size, as that one is always sizeof(sync_fence_info).
>>  - Remove the size checking in the same function and make its return type void
>>
>> Then one can simplify this loop even further :-)
>
> Sounds good to me.
>
>>
>> > +                                          size - len);
>> >
>> >                 if (ret < 0)
>> >                         goto out;
>> > @@ -540,14 +560,24 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>> >                 len += ret;
>> >         }
>> >
>> > +       if (copy_to_user((void __user *)in.sync_fence_info, fence_info, size)) {
>> > +               ret = -EFAULT;
>> > +               goto out;
>> > +       }
>> > +
>> >         info->len = len;
>> > +       info->sync_fence_info = (__u64) in.sync_fence_info;
>> Why the cast ?
>>
>> > +
>> > +no_fences:
>> > +       info->num_fences = sync_file->num_fences;
>> >
>> > -       if (copy_to_user((void __user *)arg, info, len))
>> > +       if (copy_to_user((void __user *)arg, info, sizeof(*info)))
>> Don't know if we should be returning (copying) any other information
>> but info->num_fences in case of "no_fences". In case it's not clear -
>> I'm thinking about the data we already have in in info->name and
>> info->status.
>
> Userspace might want to know all info about the sync_file but
> sync_fence_info.
>
IMHO, that does sound like a rather strange use of the API. Whichever
route one goes for, it would be nice to have the behaviour clearly
documented. Otherwise things will end up quite confusing.
Documentation can be done in separate patch/series.

Thanks
Emil

Re: [PATCH] staging/android: refactor SYNC_IOC_FILE_INFO

[Gustavo Padovan]

Hi Maarten,

2016-02-29 Maarten Lankhorst <maarten.lankhorst@linux.intel.com>:

> Op 26-02-16 om 22:00 schreef Gustavo Padovan:
> > From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> >
> > Change SYNC_IOC_FILE_INFO behaviour to avoid future API breaks and
> > optimize buffer allocation. In the new approach the ioctl needs to be called
> > twice to retrieve the array of fence_infos pointed by info->sync_fence_info.
> >
> > The first call should pass num_fences = 0, the kernel will then fill
> > info->num_fences. Userspace receives back the number of fences and
> > allocates a buffer size num_fences * sizeof(struct sync_fence_info) on
> > info->sync_fence_info.
> >
> > It then call the ioctl again passing num_fences received in info->num_fences.
> > The kernel checks if info->num_fences > 0 and if yes it fill
> > info->sync_fence_info with an array containing all fence_infos.
> >
> > info->len now represents the length of the buffer sync_fence_info points
> > to. Also, info->sync_fence_info was converted to __u64 pointer.
> >
> > An example userspace code would be:
> >
> > 	struct sync_file_info *info;
> > 	int err, size, num_fences;
> >
> > 	info = malloc(sizeof(*info));
> >
> > 	memset(info, 0, sizeof(*info));
> >
> > 	err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
> > 	num_fences = info->num_fences;
> >
> > 	if (num_fences) {
> > 		memset(info, 0, sizeof(*info));
> Would this memset still be needed if we didn't check for nulls in info->status and info->name ?
> 
> Seems to me that it could be skipped in that case.

Yes, I agree.

> > 		size = sizeof(struct sync_fence_info) * num_fences;
> > 		info->len = size;
> > 		info->num_fences = num_fences;
> > 		info->sync_fence_info = (uint64_t) calloc(num_fences,
> > 							  sizeof(struct sync_fence_info));
> >
> > 		err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
> > 	}
> >
> > v2: fix fence_info memory leak
> >
> > Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> > ---
> >  drivers/staging/android/sync.c      | 52 +++++++++++++++++++++++++++++--------
> >  drivers/staging/android/uapi/sync.h |  9 +++----
> >  2 files changed, 45 insertions(+), 16 deletions(-)
> >
> > diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
> > index dc5f382..2379f23 100644
> > --- a/drivers/staging/android/sync.c
> > +++ b/drivers/staging/android/sync.c
> > @@ -502,21 +502,22 @@ static int sync_fill_fence_info(struct fence *fence, void *data, int size)
> >  static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >  					unsigned long arg)
> >  {
> > -	struct sync_file_info *info;
> > +	struct sync_file_info in, *info;
> > +	struct sync_fence_info *fence_info = NULL;
> >  	__u32 size;
> >  	__u32 len = 0;
> = 0 unneeded.
> >  	int ret, i;
> >  
> > -	if (copy_from_user(&size, (void __user *)arg, sizeof(size)))
> > +	if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
> >  		return -EFAULT;
> >  
> > -	if (size < sizeof(struct sync_file_info))
> > -		return -EINVAL;
> > +	if (in.status || strcmp(in.name, "\0"))
> > +		return -EFAULT;
> These members always get written by the fence ioctl, I'm not sure it adds value to have them explicitly zero'd out by userspace.
> > -	if (size > 4096)
> > -		size = 4096;
> > +	if (in.num_fences && !in.sync_fence_info)
> > +		return -EFAULT;
> This check is unneeded, it will happen in the copy_to_user call anyway.
> > -	info = kzalloc(size, GFP_KERNEL);
> > +	info = kzalloc(sizeof(*info), GFP_KERNEL);
> >  	if (!info)
> >  		return -ENOMEM;
> >  
> > @@ -525,14 +526,33 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >  	if (info->status >= 0)
> >  		info->status = !info->status;
> >  
> > -	info->num_fences = sync_file->num_fences;
> > +	/*
> > +	 * Passing num_fences = 0 means that userspace want to know how
> > +	 * many fences are in the sync_file to be able to allocate a buffer to
> > +	 * fit all sync_fence_infos and call the ioctl again with the buffer
> > +	 * assigned to info->sync_fence_info. The second call pass the
> > +	 * num_fences value received in the first call.
> > +	 */
> > +	if (!in.num_fences)
> > +		goto no_fences;
> > +
> > +	size = sync_file->num_fences * sizeof(*fence_info);
> > +	if (in.len != size) {
> > +		ret = -EFAULT;
> > +		goto out;
> > +	}
> Maybe check for in.len < size, and set set to size?
> 
> 
> > -	len = sizeof(struct sync_file_info);
> > +	fence_info = kzalloc(size, GFP_KERNEL);
> > +	if (!fence_info) {
> > +		ret = -ENOMEM;
> > +		goto out;
> > +	}
> >  
> >  	for (i = 0; i < sync_file->num_fences; ++i) {
> >  		struct fence *fence = sync_file->cbs[i].fence;
> >  
> > -		ret = sync_fill_fence_info(fence, (u8 *)info + len, size - len);
> > +		ret = sync_fill_fence_info(fence, (u8 *)fence_info + len,
> > +					   size - len);
> >  
> >  		if (ret < 0)
> >  			goto out;
> > @@ -540,14 +560,24 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >  		len += ret;
> >  	}
> >  
> > +	if (copy_to_user((void __user *)in.sync_fence_info, fence_info, size)) {
> > +		ret = -EFAULT;
> > +		goto out;
> > +	}
> > +
> >  	info->len = len;
> > +	info->sync_fence_info = (__u64) in.sync_fence_info;
> > +
> > +no_fences:
> > +	info->num_fences = sync_file->num_fences;
> >  
> > -	if (copy_to_user((void __user *)arg, info, len))
> > +	if (copy_to_user((void __user *)arg, info, sizeof(*info)))
> >  		ret = -EFAULT;
> >  	else
> >  		ret = 0;
> >  
> >  out:
> > +	kfree(fence_info);
> >  	kfree(info);
> >  
> >  	return ret;
> > diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
> > index f0b41ce..9aad623 100644
> > --- a/drivers/staging/android/uapi/sync.h
> > +++ b/drivers/staging/android/uapi/sync.h
> > @@ -42,21 +42,20 @@ struct sync_fence_info {
> >  
> >  /**
> >   * struct sync_file_info - data returned from fence info ioctl
> > - * @len:	ioctl caller writes the size of the buffer its passing in.
> > - *		ioctl returns length of sync_file_info returned to
> > - *		userspace including pt_info.
> >   * @name:	name of fence
> >   * @status:	status of fence. 1: signaled 0:active <0:error
> >   * @num_fences	number of fences in the sync_file
> > + * @len:	ioctl caller writes the size of the buffer its passing in.
> > + *		ioctl returns length of all fence_infos summed.
> >   * @sync_fence_info: array of sync_fence_info for every fence in the sync_file
> >   */
> >  struct sync_file_info {
> > -	__u32	len;
> >  	char	name[32];
> >  	__s32	status;
> >  	__u32	num_fences;
> > +	__u32	len;
> >  
> > -	__u8	sync_fence_info[0];
> > +	__u64	sync_fence_info;
> >  };
> >  
> >  #define SYNC_IOC_MAGIC		'>'
> Not sure if len adds anything here, userspace knows to allocate num_fences * sizeof(struct sync_fence_info);

I think of len being useful if we decide to extend struct sync_fence_info in
the future, so we may use len to help discover the size of each
sync_fence_info. What do you think?

	Gustavo

Re: [PATCH v4 1/5] staging/android: add num_fences field to struct sync_file_info

[Dan Carpenter]

This breaks userspace.  You used to be able to figure it out from
info->len - sizeof(struct sync_file_info).

regards,
dan carpenter

Re: [PATCH v4 4/5] staging/android: refactor SYNC_IOC_FILE_INFO

[Dan Carpenter]

On Fri, Feb 26, 2016 at 03:31:46PM -0300, Gustavo Padovan wrote:
> +no_fences:
> +	info->num_fences = sync_file->num_fences;
>  
> -	if (copy_to_user((void __user *)arg, info, len))
> +	if (copy_to_user((void __user *)arg, info, sizeof(*info)))
>  		ret = -EFAULT;
>  	else
>  		ret = 0;

We need to kfree(fence_info) here.

> diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
> index f0b41ce..9aad623 100644
> --- a/drivers/staging/android/uapi/sync.h
> +++ b/drivers/staging/android/uapi/sync.h
> @@ -42,21 +42,20 @@ struct sync_fence_info {
>  
>  /**
>   * struct sync_file_info - data returned from fence info ioctl
> - * @len:	ioctl caller writes the size of the buffer its passing in.
> - *		ioctl returns length of sync_file_info returned to
> - *		userspace including pt_info.
>   * @name:	name of fence
>   * @status:	status of fence. 1: signaled 0:active <0:error
>   * @num_fences	number of fences in the sync_file
> + * @len:	ioctl caller writes the size of the buffer its passing in.
> + *		ioctl returns length of all fence_infos summed.
>   * @sync_fence_info: array of sync_fence_info for every fence in the sync_file

The documentation needs updating.

>   */
>  struct sync_file_info {
> -	__u32	len;
>  	char	name[32];
>  	__s32	status;
>  	__u32	num_fences;
> +	__u32	len;
>  
> -	__u8	sync_fence_info[0];
> +	__u64	sync_fence_info;
>  };


regards,
dan carpenter

Re: [PATCH] staging/android: refactor SYNC_IOC_FILE_INFO

[Maarten Lankhorst]

Op 29-02-16 om 23:08 schreef Gustavo Padovan:
> Hi Maarten,
>
> 2016-02-29 Maarten Lankhorst <maarten.lankhorst@linux.intel.com>:
>
>> Op 26-02-16 om 22:00 schreef Gustavo Padovan:
>>> From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>>>
>>> Change SYNC_IOC_FILE_INFO behaviour to avoid future API breaks and
>>> optimize buffer allocation. In the new approach the ioctl needs to be called
>>> twice to retrieve the array of fence_infos pointed by info->sync_fence_info.
>>>
>>> The first call should pass num_fences = 0, the kernel will then fill
>>> info->num_fences. Userspace receives back the number of fences and
>>> allocates a buffer size num_fences * sizeof(struct sync_fence_info) on
>>> info->sync_fence_info.
>>>
>>> It then call the ioctl again passing num_fences received in info->num_fences.
>>> The kernel checks if info->num_fences > 0 and if yes it fill
>>> info->sync_fence_info with an array containing all fence_infos.
>>>
>>> info->len now represents the length of the buffer sync_fence_info points
>>> to. Also, info->sync_fence_info was converted to __u64 pointer.
>>>
>>> An example userspace code would be:
>>>
>>> 	struct sync_file_info *info;
>>> 	int err, size, num_fences;
>>>
>>> 	info = malloc(sizeof(*info));
>>>
>>> 	memset(info, 0, sizeof(*info));
>>>
>>> 	err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
>>> 	num_fences = info->num_fences;
>>>
>>> 	if (num_fences) {
>>> 		memset(info, 0, sizeof(*info));
>> Would this memset still be needed if we didn't check for nulls in info->status and info->name ?
>>
>> Seems to me that it could be skipped in that case.
> Yes, I agree.
>
>>> 		size = sizeof(struct sync_fence_info) * num_fences;
>>> 		info->len = size;
>>> 		info->num_fences = num_fences;
>>> 		info->sync_fence_info = (uint64_t) calloc(num_fences,
>>> 							  sizeof(struct sync_fence_info));
>>>
>>> 		err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
>>> 	}
>>>
>>> v2: fix fence_info memory leak
>>>
>>> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>>> ---
>>>  drivers/staging/android/sync.c      | 52 +++++++++++++++++++++++++++++--------
>>>  drivers/staging/android/uapi/sync.h |  9 +++----
>>>  2 files changed, 45 insertions(+), 16 deletions(-)
>>>
>>> diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
>>> index dc5f382..2379f23 100644
>>> --- a/drivers/staging/android/sync.c
>>> +++ b/drivers/staging/android/sync.c
>>> @@ -502,21 +502,22 @@ static int sync_fill_fence_info(struct fence *fence, void *data, int size)
>>>  static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>>>  					unsigned long arg)
>>>  {
>>> -	struct sync_file_info *info;
>>> +	struct sync_file_info in, *info;
>>> +	struct sync_fence_info *fence_info = NULL;
>>>  	__u32 size;
>>>  	__u32 len = 0;
>> = 0 unneeded.
>>>  	int ret, i;
>>>  
>>> -	if (copy_from_user(&size, (void __user *)arg, sizeof(size)))
>>> +	if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
>>>  		return -EFAULT;
>>>  
>>> -	if (size < sizeof(struct sync_file_info))
>>> -		return -EINVAL;
>>> +	if (in.status || strcmp(in.name, "\0"))
>>> +		return -EFAULT;
>> These members always get written by the fence ioctl, I'm not sure it adds value to have them explicitly zero'd out by userspace.
>>> -	if (size > 4096)
>>> -		size = 4096;
>>> +	if (in.num_fences && !in.sync_fence_info)
>>> +		return -EFAULT;
>> This check is unneeded, it will happen in the copy_to_user call anyway.
>>> -	info = kzalloc(size, GFP_KERNEL);
>>> +	info = kzalloc(sizeof(*info), GFP_KERNEL);
>>>  	if (!info)
>>>  		return -ENOMEM;
>>>  
>>> @@ -525,14 +526,33 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>>>  	if (info->status >= 0)
>>>  		info->status = !info->status;
>>>  
>>> -	info->num_fences = sync_file->num_fences;
>>> +	/*
>>> +	 * Passing num_fences = 0 means that userspace want to know how
>>> +	 * many fences are in the sync_file to be able to allocate a buffer to
>>> +	 * fit all sync_fence_infos and call the ioctl again with the buffer
>>> +	 * assigned to info->sync_fence_info. The second call pass the
>>> +	 * num_fences value received in the first call.
>>> +	 */
>>> +	if (!in.num_fences)
>>> +		goto no_fences;
>>> +
>>> +	size = sync_file->num_fences * sizeof(*fence_info);
>>> +	if (in.len != size) {
>>> +		ret = -EFAULT;
>>> +		goto out;
>>> +	}
>> Maybe check for in.len < size, and set set to size?
>>
>>
>>> -	len = sizeof(struct sync_file_info);
>>> +	fence_info = kzalloc(size, GFP_KERNEL);
>>> +	if (!fence_info) {
>>> +		ret = -ENOMEM;
>>> +		goto out;
>>> +	}
>>>  
>>>  	for (i = 0; i < sync_file->num_fences; ++i) {
>>>  		struct fence *fence = sync_file->cbs[i].fence;
>>>  
>>> -		ret = sync_fill_fence_info(fence, (u8 *)info + len, size - len);
>>> +		ret = sync_fill_fence_info(fence, (u8 *)fence_info + len,
>>> +					   size - len);
>>>  
>>>  		if (ret < 0)
>>>  			goto out;
>>> @@ -540,14 +560,24 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
>>>  		len += ret;
>>>  	}
>>>  
>>> +	if (copy_to_user((void __user *)in.sync_fence_info, fence_info, size)) {
>>> +		ret = -EFAULT;
>>> +		goto out;
>>> +	}
>>> +
>>>  	info->len = len;
>>> +	info->sync_fence_info = (__u64) in.sync_fence_info;
>>> +
>>> +no_fences:
>>> +	info->num_fences = sync_file->num_fences;
>>>  
>>> -	if (copy_to_user((void __user *)arg, info, len))
>>> +	if (copy_to_user((void __user *)arg, info, sizeof(*info)))
>>>  		ret = -EFAULT;
>>>  	else
>>>  		ret = 0;
>>>  
>>>  out:
>>> +	kfree(fence_info);
>>>  	kfree(info);
>>>  
>>>  	return ret;
>>> diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
>>> index f0b41ce..9aad623 100644
>>> --- a/drivers/staging/android/uapi/sync.h
>>> +++ b/drivers/staging/android/uapi/sync.h
>>> @@ -42,21 +42,20 @@ struct sync_fence_info {
>>>  
>>>  /**
>>>   * struct sync_file_info - data returned from fence info ioctl
>>> - * @len:	ioctl caller writes the size of the buffer its passing in.
>>> - *		ioctl returns length of sync_file_info returned to
>>> - *		userspace including pt_info.
>>>   * @name:	name of fence
>>>   * @status:	status of fence. 1: signaled 0:active <0:error
>>>   * @num_fences	number of fences in the sync_file
>>> + * @len:	ioctl caller writes the size of the buffer its passing in.
>>> + *		ioctl returns length of all fence_infos summed.
>>>   * @sync_fence_info: array of sync_fence_info for every fence in the sync_file
>>>   */
>>>  struct sync_file_info {
>>> -	__u32	len;
>>>  	char	name[32];
>>>  	__s32	status;
>>>  	__u32	num_fences;
>>> +	__u32	len;
>>>  
>>> -	__u8	sync_fence_info[0];
>>> +	__u64	sync_fence_info;
>>>  };
>>>  
>>>  #define SYNC_IOC_MAGIC		'>'
>> Not sure if len adds anything here, userspace knows to allocate num_fences * sizeof(struct sync_fence_info);
> I think of len being useful if we decide to extend struct sync_fence_info in
> the future, so we may use len to help discover the size of each
> sync_fence_info. What do you think?
>
I don't think you could extend it arbitrarily, you could make userspace pass a flag to indicate the struct is extended, so kernel space can choose
whether to use the bigger size struct or not.

something like sync_file_info.flags = FENCE_INFO_V2; -- kernel can reject with -EINVAL if unsupported, or fill in a v2 struct.

~Maarten

Re: [PATCH v4 1/5] staging/android: add num_fences field to struct sync_file_info

[Gustavo Padovan]

2016-03-01 Dan Carpenter <dan.carpenter@oracle.com>:

> This breaks userspace.  You used to be able to figure it out from
> info->len - sizeof(struct sync_file_info).

It does. We are breaking this on purpose to clean up the API for
de-staging.

	Gustavo

Re: [PATCH] staging/android: refactor SYNC_IOC_FILE_INFO

[Gustavo Padovan]

Hi Maarten,

2016-03-01 Maarten Lankhorst <maarten.lankhorst@linux.intel.com>:

> Op 29-02-16 om 23:08 schreef Gustavo Padovan:
> > Hi Maarten,
> >
> > 2016-02-29 Maarten Lankhorst <maarten.lankhorst@linux.intel.com>:
> >
> >> Op 26-02-16 om 22:00 schreef Gustavo Padovan:
> >>> From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> >>>
> >>> Change SYNC_IOC_FILE_INFO behaviour to avoid future API breaks and
> >>> optimize buffer allocation. In the new approach the ioctl needs to be called
> >>> twice to retrieve the array of fence_infos pointed by info->sync_fence_info.
> >>>
> >>> The first call should pass num_fences = 0, the kernel will then fill
> >>> info->num_fences. Userspace receives back the number of fences and
> >>> allocates a buffer size num_fences * sizeof(struct sync_fence_info) on
> >>> info->sync_fence_info.
> >>>
> >>> It then call the ioctl again passing num_fences received in info->num_fences.
> >>> The kernel checks if info->num_fences > 0 and if yes it fill
> >>> info->sync_fence_info with an array containing all fence_infos.
> >>>
> >>> info->len now represents the length of the buffer sync_fence_info points
> >>> to. Also, info->sync_fence_info was converted to __u64 pointer.
> >>>
> >>> An example userspace code would be:
> >>>
> >>> 	struct sync_file_info *info;
> >>> 	int err, size, num_fences;
> >>>
> >>> 	info = malloc(sizeof(*info));
> >>>
> >>> 	memset(info, 0, sizeof(*info));
> >>>
> >>> 	err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
> >>> 	num_fences = info->num_fences;
> >>>
> >>> 	if (num_fences) {
> >>> 		memset(info, 0, sizeof(*info));
> >> Would this memset still be needed if we didn't check for nulls in info->status and info->name ?
> >>
> >> Seems to me that it could be skipped in that case.
> > Yes, I agree.
> >
> >>> 		size = sizeof(struct sync_fence_info) * num_fences;
> >>> 		info->len = size;
> >>> 		info->num_fences = num_fences;
> >>> 		info->sync_fence_info = (uint64_t) calloc(num_fences,
> >>> 							  sizeof(struct sync_fence_info));
> >>>
> >>> 		err = ioctl(fd, SYNC_IOC_FILE_INFO, info);
> >>> 	}
> >>>
> >>> v2: fix fence_info memory leak
> >>>
> >>> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> >>> ---
> >>>  drivers/staging/android/sync.c      | 52 +++++++++++++++++++++++++++++--------
> >>>  drivers/staging/android/uapi/sync.h |  9 +++----
> >>>  2 files changed, 45 insertions(+), 16 deletions(-)
> >>>
> >>> diff --git a/drivers/staging/android/sync.c b/drivers/staging/android/sync.c
> >>> index dc5f382..2379f23 100644
> >>> --- a/drivers/staging/android/sync.c
> >>> +++ b/drivers/staging/android/sync.c
> >>> @@ -502,21 +502,22 @@ static int sync_fill_fence_info(struct fence *fence, void *data, int size)
> >>>  static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >>>  					unsigned long arg)
> >>>  {
> >>> -	struct sync_file_info *info;
> >>> +	struct sync_file_info in, *info;
> >>> +	struct sync_fence_info *fence_info = NULL;
> >>>  	__u32 size;
> >>>  	__u32 len = 0;
> >> = 0 unneeded.
> >>>  	int ret, i;
> >>>  
> >>> -	if (copy_from_user(&size, (void __user *)arg, sizeof(size)))
> >>> +	if (copy_from_user(&in, (void __user *)arg, sizeof(*info)))
> >>>  		return -EFAULT;
> >>>  
> >>> -	if (size < sizeof(struct sync_file_info))
> >>> -		return -EINVAL;
> >>> +	if (in.status || strcmp(in.name, "\0"))
> >>> +		return -EFAULT;
> >> These members always get written by the fence ioctl, I'm not sure it adds value to have them explicitly zero'd out by userspace.
> >>> -	if (size > 4096)
> >>> -		size = 4096;
> >>> +	if (in.num_fences && !in.sync_fence_info)
> >>> +		return -EFAULT;
> >> This check is unneeded, it will happen in the copy_to_user call anyway.
> >>> -	info = kzalloc(size, GFP_KERNEL);
> >>> +	info = kzalloc(sizeof(*info), GFP_KERNEL);
> >>>  	if (!info)
> >>>  		return -ENOMEM;
> >>>  
> >>> @@ -525,14 +526,33 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >>>  	if (info->status >= 0)
> >>>  		info->status = !info->status;
> >>>  
> >>> -	info->num_fences = sync_file->num_fences;
> >>> +	/*
> >>> +	 * Passing num_fences = 0 means that userspace want to know how
> >>> +	 * many fences are in the sync_file to be able to allocate a buffer to
> >>> +	 * fit all sync_fence_infos and call the ioctl again with the buffer
> >>> +	 * assigned to info->sync_fence_info. The second call pass the
> >>> +	 * num_fences value received in the first call.
> >>> +	 */
> >>> +	if (!in.num_fences)
> >>> +		goto no_fences;
> >>> +
> >>> +	size = sync_file->num_fences * sizeof(*fence_info);
> >>> +	if (in.len != size) {
> >>> +		ret = -EFAULT;
> >>> +		goto out;
> >>> +	}
> >> Maybe check for in.len < size, and set set to size?
> >>
> >>
> >>> -	len = sizeof(struct sync_file_info);
> >>> +	fence_info = kzalloc(size, GFP_KERNEL);
> >>> +	if (!fence_info) {
> >>> +		ret = -ENOMEM;
> >>> +		goto out;
> >>> +	}
> >>>  
> >>>  	for (i = 0; i < sync_file->num_fences; ++i) {
> >>>  		struct fence *fence = sync_file->cbs[i].fence;
> >>>  
> >>> -		ret = sync_fill_fence_info(fence, (u8 *)info + len, size - len);
> >>> +		ret = sync_fill_fence_info(fence, (u8 *)fence_info + len,
> >>> +					   size - len);
> >>>  
> >>>  		if (ret < 0)
> >>>  			goto out;
> >>> @@ -540,14 +560,24 @@ static long sync_file_ioctl_fence_info(struct sync_file *sync_file,
> >>>  		len += ret;
> >>>  	}
> >>>  
> >>> +	if (copy_to_user((void __user *)in.sync_fence_info, fence_info, size)) {
> >>> +		ret = -EFAULT;
> >>> +		goto out;
> >>> +	}
> >>> +
> >>>  	info->len = len;
> >>> +	info->sync_fence_info = (__u64) in.sync_fence_info;
> >>> +
> >>> +no_fences:
> >>> +	info->num_fences = sync_file->num_fences;
> >>>  
> >>> -	if (copy_to_user((void __user *)arg, info, len))
> >>> +	if (copy_to_user((void __user *)arg, info, sizeof(*info)))
> >>>  		ret = -EFAULT;
> >>>  	else
> >>>  		ret = 0;
> >>>  
> >>>  out:
> >>> +	kfree(fence_info);
> >>>  	kfree(info);
> >>>  
> >>>  	return ret;
> >>> diff --git a/drivers/staging/android/uapi/sync.h b/drivers/staging/android/uapi/sync.h
> >>> index f0b41ce..9aad623 100644
> >>> --- a/drivers/staging/android/uapi/sync.h
> >>> +++ b/drivers/staging/android/uapi/sync.h
> >>> @@ -42,21 +42,20 @@ struct sync_fence_info {
> >>>  
> >>>  /**
> >>>   * struct sync_file_info - data returned from fence info ioctl
> >>> - * @len:	ioctl caller writes the size of the buffer its passing in.
> >>> - *		ioctl returns length of sync_file_info returned to
> >>> - *		userspace including pt_info.
> >>>   * @name:	name of fence
> >>>   * @status:	status of fence. 1: signaled 0:active <0:error
> >>>   * @num_fences	number of fences in the sync_file
> >>> + * @len:	ioctl caller writes the size of the buffer its passing in.
> >>> + *		ioctl returns length of all fence_infos summed.
> >>>   * @sync_fence_info: array of sync_fence_info for every fence in the sync_file
> >>>   */
> >>>  struct sync_file_info {
> >>> -	__u32	len;
> >>>  	char	name[32];
> >>>  	__s32	status;
> >>>  	__u32	num_fences;
> >>> +	__u32	len;
> >>>  
> >>> -	__u8	sync_fence_info[0];
> >>> +	__u64	sync_fence_info;
> >>>  };
> >>>  
> >>>  #define SYNC_IOC_MAGIC		'>'
> >> Not sure if len adds anything here, userspace knows to allocate num_fences * sizeof(struct sync_fence_info);
> > I think of len being useful if we decide to extend struct sync_fence_info in
> > the future, so we may use len to help discover the size of each
> > sync_fence_info. What do you think?
> >
> I don't think you could extend it arbitrarily, you could make userspace pass a flag to indicate the struct is extended, so kernel space can choose
> whether to use the bigger size struct or not.
> 
> something like sync_file_info.flags = FENCE_INFO_V2; -- kernel can reject with -EINVAL if unsupported, or fill in a v2 struct.

Fair enough, I'll just remove len then.

	Gustavo