Skip to contentSkip to conversation

Improve the configfs read and write iterators further

PATCH In review Bart Van Assche 3 participants 17 messages Jul 23, 2021 – Jul 29, 2021

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover Improve the configfs read and write iterators further

Hi Christoph,

This patch series addresses Linus' feedback on my most recent configfs patch, adds unit tests for configfs and also includes one functional change that I came up with while writing these unit tests. I will leave it to you to decide whether to submit these patches for inclusion in kernel v5.14 or v5.15.

Thanks,

Bart.

Bart Van Assche (4):
configfs: Rework the overflow check in fill_write_buffer()
configfs: Fix writing at a non-zero offset
kunit: Add support for suite initialization and cleanup
configfs: Add unit tests

 fs/configfs/Kconfig         |   8 +
 fs/configfs/Makefile        |   2 +
 fs/configfs/configfs-test.c | 369 ++++++++++++++++++++++++++++++++++++
 fs/configfs/file.c          |  17 +-
 include/kunit/test.h        |   4 +
 lib/kunit/test.c            |  14 ++
 6 files changed, 406 insertions(+), 8 deletions(-)
 create mode 100644 fs/configfs/configfs-test.c
#1 lore raw

PATCH 1/4 configfs: Rework the overflow check in fill_write_buffer()

Change 'if (SIMPLE_ATTR_SIZE - 1 - pos <= 0)' into
'if (pos >= SIMPLE_ATTR_SIZE - 1)'. Change the data type of 'to_copy' from long long (loff_t) into int. Do not check whether pos < 0 since rw_verify_area() checks this for us. As one can see on
https://lore.kernel.org/lkml/CAHk-=wjuDBQdUvaO=XaptgmvE_qeg_EuZjsUZf2vVoXPUMgAvg@mail.gmail.com/
these changes have been requested by Linus Torvalds.

Cc Bodo Stroesser <[email protected]>
Cc Martin K. Petersen <[email protected]>
Cc Yanko Kaneti <[email protected]>
Cc Brendan Higgins <[email protected]>
---
 fs/configfs/file.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
fs/configfs/file.c +3 -4
index 5a0be9985bae..8121bb1b2121 100644
@@ -181,8 +181,7 @@ static ssize_t configfs_bin_read_iter(struct kiocb *iocb, struct iov_iter *to)
 static int fill_write_buffer(struct configfs_buffer *buffer, loff_t pos,
 			     struct iov_iter *from)
 {
-	loff_t to_copy;
-	int copied;
+	int to_copy, copied;
 	u8 *to;
 
 	if (!buffer->page)
@@ -190,9 +189,9 @@ static int fill_write_buffer(struct configfs_buffer *buffer, loff_t pos,
 	if (!buffer->page)
 		return -ENOMEM;
 
-	to_copy = SIMPLE_ATTR_SIZE - 1 - pos;
-	if (to_copy <= 0)
+	if (pos >= SIMPLE_ATTR_SIZE - 1)
 		return 0;
+	to_copy = SIMPLE_ATTR_SIZE - 1 - pos;
 	to = buffer->page + pos;
 	copied = copy_from_iter(to, to_copy, from);
 	buffer->needs_read_fill = 1;
#2 lore raw

PATCH 2/4 configfs: Fix writing at a non-zero offset

In reply to Bart Van Assche

Store data at the right offset when writing with a non-zero offset. I think this patch fixes behavior introduced in the initial configfs commit. From a source code comment in the initial configfs commit: "There is no easy way for us to know if userspace is only doing a partial write, so we don't support them."

Cc Bodo Stroesser <[email protected]>
Cc Martin K. Petersen <[email protected]>
Cc Yanko Kaneti <[email protected]>
Cc Brendan Higgins <[email protected]>
---
 fs/configfs/file.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
fs/configfs/file.c +6 -4
index 8121bb1b2121..3b2ddc2e5d42 100644
@@ -226,14 +226,16 @@ static ssize_t configfs_write_iter(struct kiocb *iocb, struct iov_iter *from)
 {
 	struct file *file = iocb->ki_filp;
 	struct configfs_buffer *buffer = file->private_data;
-	ssize_t len;
+	int len, written = 0;
 
 	mutex_lock(&buffer->mutex);
 	len = fill_write_buffer(buffer, iocb->ki_pos, from);
 	if (len > 0)
-		len = flush_write_buffer(file, buffer, len);
-	if (len > 0)
-		iocb->ki_pos += len;
+		written = flush_write_buffer(file, buffer, iocb->ki_pos + len);
+	if (written > 0) {
+		len = max_t(int, 0, written - iocb->ki_pos);
+		iocb->ki_pos = written;
+	}
 	mutex_unlock(&buffer->mutex);
 	return len;
 }
#3 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Quoted 2 lines · Bart Van Assche

Store data at the right offset when writing with a non-zero offset. I think this patch fixes behavior introduced in the initial configfs commit.

Unfortunately I think it does not.

Let's say user writes 5 times to configfs file while keeping it open. On every write() call it writes 1 character only, e.g. first "A", then "B", ...

The original code before the changes 5 times called flush_write_buffer for the strings "A\0", "B\0", ... (with the '\0' not included in the count parameter, so count is 1 always, which is the length of the last write).

Since commit
420405ecde06 "configfs: fix the read and write iterators"
flush_write_buffer is called for the strings "A\0", "AB\0", "ABC\0", ... but count still is 1.

This fix changes just the count parameter, so it now is 1, 2, 3, ... in our example. But it still sends on every call to flush_write_buffer all the write data gathered since last open().

I think, a simple way to restore the original behavior would be to revert the part of commit 420405ecde06 that changed write routines.

Bodo

From a

Quoted 39 lines

source code comment in the initial configfs commit: "There is no easy way for us to know if userspace is only doing a partial write, so we don't support them."

Cc Bodo Stroesser <[email protected]>
Cc Martin K. Petersen <[email protected]>
Cc Yanko Kaneti <[email protected]>
Cc Brendan Higgins <[email protected]>
---
  fs/configfs/file.c | 10 ++++++----
  1 file changed, 6 insertions(+), 4 deletions(-)
fs/configfs/file.c +6 -4
index 8121bb1b2121..3b2ddc2e5d42 100644
@@ -226,14 +226,16 @@ static ssize_t configfs_write_iter(struct kiocb *iocb, struct iov_iter *from)
  {
  	struct file *file = iocb->ki_filp;
  	struct configfs_buffer *buffer = file->private_data;
-	ssize_t len;
+	int len, written = 0;
 
  	mutex_lock(&buffer->mutex);
  	len = fill_write_buffer(buffer, iocb->ki_pos, from);
  	if (len > 0)
-		len = flush_write_buffer(file, buffer, len);
-	if (len > 0)
-		iocb->ki_pos += len;
+		written = flush_write_buffer(file, buffer, iocb->ki_pos + len);
+	if (written > 0) {
+		len = max_t(int, 0, written - iocb->ki_pos);
+		iocb->ki_pos = written;
+	}
  	mutex_unlock(&buffer->mutex);
  	return len;
  }
#4 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Quoted 7 lines · Bodo Stroesser

On 23.07.21 23:23, Bart Van Assche wrote:
Let's say user writes 5 times to configfs file while keeping it open. On every write() call it writes 1 character only, e.g. first "A", then "B", ...

The original code before the changes 5 times called flush_write_buffer for the strings "A\0", "B\0", ... (with the '\0' not included in the count parameter, so count is 1 always, which is the length of the last write).

Isn't that behavior a severe violation of how POSIX specifies that the write() system call should be implemented?

Thanks,

Bart.

#5 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Quoted 10 lines · Bart Van Assche
Quoted 7 lines · Bodo Stroesser

On 23.07.21 23:23, Bart Van Assche wrote:
Let's say user writes 5 times to configfs file while keeping it open. On every write() call it writes 1 character only, e.g. first "A", then "B", ...

The original code before the changes 5 times called flush_write_buffer for the strings "A\0", "B\0", ... (with the '\0' not included in the count parameter, so count is 1 always, which is the length of the last write).

Isn't that behavior a severe violation of how POSIX specifies that the write() system call should be implemented?

Hmm. I'm not sure which detail should violate POSIX spec? Is there any definition how data should be flushed from buffer internally? (I'm by far not a POSIX expert!)

I would rather say the new behavior, to call flush_write_buffer during the first write() for the data of that write, and then on the second write to call flush_write_buffer for the concatenated data of the first and the second write, could be a violation of POSIX, because the one times written data of the first write is flushed twice.

I don't like the idea of breaking the "one write, one flush" principle that was implemented before. The old comment:
"There is no easy way for us to know if userspace is only doing a partial write, so we don't support them. We expect the entire buffer to come on the first write."
as I interpret it, makes clear that configfs code has to work according to that principle. (Or even block all but the first write, but that would even more break compatibility to old implementation.)

Thank you,
Bodo

#6 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Quoted 22 lines · Bodo Stroesser
Quoted 10 lines · Bart Van Assche
Quoted 7 lines · Bodo Stroesser

On 23.07.21 23:23, Bart Van Assche wrote:
Let's say user writes 5 times to configfs file while keeping it open. On every write() call it writes 1 character only, e.g. first "A", then "B", ...

The original code before the changes 5 times called flush_write_buffer for the strings "A\0", "B\0", ... (with the '\0' not included in the count parameter, so count is 1 always, which is the length of the last write).

Isn't that behavior a severe violation of how POSIX specifies that the write() system call should be implemented?

Hmm. I'm not sure which detail should violate POSIX spec? Is there any definition how data should be flushed from buffer internally? (I'm by far not a POSIX expert!)

I would rather say the new behavior, to call flush_write_buffer during the first write() for the data of that write, and then on the second write to call flush_write_buffer for the concatenated data of the first and the second write, could be a violation of POSIX, because the one times written data of the first write is flushed twice.

Quoted 8 lines

I don't like the idea of breaking the "one write, one flush" principle that was implemented before. The old comment:
"There is no easy way for us to know if userspace is only doing a partial write, so we don't support them. We expect the entire buffer to come on the first write."
as I interpret it, makes clear that configfs code has to work according to that principle. (Or even block all but the first write, but that would even more break compatibility to old implementation.)

Hi Bodo,

The private email that you sent me made it clear that you would like to keep the behavior from kernel 5.13. That means passing "A\0", "B\0", ... to the configfs store callback function if "AB..." is witten one byte at a time. What is not clear to me is how a store callback with argument "B\0" can know at which offset that write happened? From <linux/configfs.h> (I have added argument names):

	ssize_t (*store)(struct config_item *item, const char *page,
                         size_t count);

My understanding of the POSIX specification [1] is that writes should happen at the proper offset. If user space software writes "A" at offset 0 and "B" at offset 1 then the string "AB" should be passed to the configfs store callback.

Regarding the "action" attribute from your tcmu patch, how about checking the last character of the string written into that attribute instead of the first character? Would that be sufficient to write twice into that attribute without having to call close() and open() between the two write actions?

To me the following comment: "There is no easy way for us to know if userspace is only doing a partial write, so we don't support them. We expect the entire buffer to come on the first write." means that writing "ABCD" by first writing "AB" and next "CD" will trigger two item->store() calls. Triggering a single item->store() call for partial writes is not supported.

Thanks,

Bart.

[1] https://pubs.opengroup.org/onlinepubs/9699919799/

#7 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Quoted 32 lines · Bart Van Assche
Quoted 22 lines · Bodo Stroesser
Quoted 10 lines · Bart Van Assche
Quoted 7 lines · Bodo Stroesser

On 23.07.21 23:23, Bart Van Assche wrote:
Let's say user writes 5 times to configfs file while keeping it open. On every write() call it writes 1 character only, e.g. first "A", then "B", ...

The original code before the changes 5 times called flush_write_buffer for the strings "A\0", "B\0", ... (with the '\0' not included in the count parameter, so count is 1 always, which is the length of the last write).

Isn't that behavior a severe violation of how POSIX specifies that the write() system call should be implemented?

Hmm. I'm not sure which detail should violate POSIX spec? Is there any definition how data should be flushed from buffer internally? (I'm by far not a POSIX expert!)

I would rather say the new behavior, to call flush_write_buffer during the first write() for the data of that write, and then on the second write to call flush_write_buffer for the concatenated data of the first and the second write, could be a violation of POSIX, because the one times written data of the first write is flushed twice.

Quoted 8 lines

I don't like the idea of breaking the "one write, one flush" principle that was implemented before. The old comment:
"There is no easy way for us to know if userspace is only doing a partial write, so we don't support them. We expect the entire buffer to come on the first write."
as I interpret it, makes clear that configfs code has to work according to that principle. (Or even block all but the first write, but that would even more break compatibility to old implementation.)

Quoted 6 lines

Hi Bodo,

The private email that you sent me made it clear that you would like to keep the behavior from kernel 5.13. That means passing "A\0", "B\0", ... to the configfs store callback function if "AB..." is witten one byte at a time. What is not clear to me is how a store callback with argument "B\0" can know at which offset that write happened? From <linux/configfs.h> (I have added argument names):

    ssize_t (*store)(struct config_item *item, const char *page, size_t count);

It does not know. It simply handles it as two separate store actions. One could say, both start from offset 0.

Quoted 2 lines

My understanding of the POSIX specification [1] is that writes should happen at the proper offset. If user space software writes "A" at offset 0 and "B" at offset 1 then the string "AB" should be passed to the configfs store callback.

The comment says, that such a concatenation is not supported. To add such a support, we would have to buffer all writes and then have a criterion that triggers the flush_write_buffer. For example that could be done on close(). But that would also mean, that bad result from store handler could be reported by close only. And it would mean, that again the behavior changes, in that the new SW allows one store action only after one open(). You have to close and re-open before you can start a new store action.

To me it looks strange to write again all previous data from the beginning at each new write. So I think this is not a good solution.

Quoted 2 lines

Regarding the "action" attribute from your tcmu patch, how about checking the last character of the string written into that attribute instead of the first character? Would that be sufficient to write twice into that attribute without having to call close() and open() between the two write actions?

I'm not sure I understand what you mean. If userspace writes a string byte by byte or in pieces of other sizes, would you still gather data in the file's buffer and call flush_write_buffer on each write with all the data gathered up to and including the current write?

Of so, do you want the store handler to detect the end of the string, e.g by searching for '\n', and discard the write if not found? That would not work well, because after the store handler detected the '\n', during the next write it would get the same string again plus what was added by the new write. Store handler would have to know, how much of the entire buffer content it already had seen. After a couple of writes we would even run out of buffer. So again close and re-open is needed. After close and re-open, how does the store handler know, that the buffer now is re-started from the beginning?

The new behavior can also cause trouble with existing store handlers. Example:
The tcmu attribute files cmd_time_out and qfull_time_out just take a string containing the decimal formatted number of seconds of the timeout. Each number up to now had to be transferred in a single write. Assume the old value is 30 and we want to change to 19. If userspace writes byte by byte, you end up calling
store(item, "1\0", 1) and then
store(item, "19\9", 2).
If these quick changes do not cause trouble in tcmu's scsi cmd handling, then think what happens, if userspace is interrupted between the two writes. Allowing to split the writes cause a loss of "atomicity".

Quoted 2 lines

To me the following comment: "There is no easy way for us to know if userspace is only doing a partial write, so we don't support them. We expect the entire buffer to come on the first write." means that writing "ABCD" by first writing "AB" and next "CD" will trigger two item->store() calls. Triggering a single item->store() call for partial writes is not supported.

Exactly. So IMHO we don't need to handle any offsets during write processing, since for every write we again start at offset 0. (We just add the trailing '\0' - not in count - to ease store handler's work.)

Thank you,
Bodo

#8 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Quoted 4 lines · Bodo Stroesser

The new behavior can also cause trouble with existing store handlers. Example:
The tcmu attribute files cmd_time_out and qfull_time_out just take a string containing the decimal formatted number of seconds of the

Quoted 8 lines

timeout. Each number up to now had to be transferred in a single write. Assume the old value is 30 and we want to change to 19. If userspace writes byte by byte, you end up calling
store(item, "1\0", 1) and then
store(item, "19\9", 2).
If these quick changes do not cause trouble in tcmu's scsi cmd handling, then think what happens, if userspace is interrupted between the two writes. Allowing to split the writes cause a loss of "atomicity".

From Documentation/filesystems/configfs.rst, for normal attributes:
"Configfs expects write(2) to store the entire buffer at once." In other words, the behavior for partial writes is undocumented. My changes preserve the behavior if a buffer is written in its entirety. I do not agree that my changes can cause trouble for existing store handlers.

Bart.

#9 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Quoted 14 lines · Bart Van Assche
Quoted 4 lines · Bodo Stroesser

The new behavior can also cause trouble with existing store handlers. Example:
The tcmu attribute files cmd_time_out and qfull_time_out just take a string containing the decimal formatted number of seconds of the

Quoted 8 lines

timeout. Each number up to now had to be transferred in a single write. Assume the old value is 30 and we want to change to 19. If userspace writes byte by byte, you end up calling
store(item, "1\0", 1) and then
store(item, "19\9", 2).
If these quick changes do not cause trouble in tcmu's scsi cmd handling, then think what happens, if userspace is interrupted between the two writes. Allowing to split the writes cause a loss of "atomicity".

Quoted 6 lines

From Documentation/filesystems/configfs.rst, for normal attributes:
"Configfs expects write(2) to store the entire buffer at once." In other words, the behavior for partial writes is undocumented. My changes preserve the behavior if a buffer is written in its entirety. I do not agree that my changes can cause trouble for existing store handlers.

I agree. I was not precise.

What I meant is, that changing the source code in such a way, that writing a buffer in multiple writes works in general, could cause trouble in case userspace uses this.

But for special syscall sequences your changes still change the result on existing configfs files. Example:

1) userspace program opens qfull_time_out
2) userspace program writes "90", count=2 to set timeout to 90 sec
3) userspace again wants to change timeout, so it writes "55", count=2

Before the changes we end up with timeout being 55 seconds. After the change - due to data gathering - we finally have timeout 9055 seconds.

BR,
Bodo

#10 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Quoted 7 lines · Bodo Stroesser

What I meant is, that changing the source code in such a way, that writing a buffer in multiple writes works in general, could cause trouble in case userspace uses this.

But for special syscall sequences your changes still change the result on existing configfs files. Example:

Quoted 6 lines

1) userspace program opens qfull_time_out
2) userspace program writes "90", count=2 to set timeout to 90 sec
3) userspace again wants to change timeout, so it writes "55", count=2

Before the changes we end up with timeout being 55 seconds. After the change - due to data gathering - we finally have timeout 9055 seconds.

Hi Bodo,

How about replacing patches 1 and 2 from this series with the patch below? Do you agree that this patch is sufficient to restore the behavior from kernel v5.13 and before?

Thanks,

Bart.

Subject: [PATCH 1/3] configfs: Restore the kernel v5.13 text attribute write behavior

Instead of writing at the offset specified by the write() system call, always write at offset zero.

Cc Bodo Stroesser <[email protected]>
Cc Martin K. Petersen <[email protected]>
Cc Yanko Kaneti <[email protected]>
Cc Brendan Higgins <[email protected]>
---
 fs/configfs/file.c | 20 ++++++++------------
 1 file changed, 8 insertions(+), 12 deletions(-)
fs/configfs/file.c +8 -12
index 5a0be9985bae..8adf6250b207 100644
@@ -177,12 +177,11 @@ static ssize_t configfs_bin_read_iter(struct kiocb *iocb, struct iov_iter *to)
 	return retval;
 }
 
-/* Fill [buffer, buffer + pos) with data coming from @from. */
-static int fill_write_buffer(struct configfs_buffer *buffer, loff_t pos,
+/* Fill @buffer with data coming from @from. */
+static int fill_write_buffer(struct configfs_buffer *buffer,
 			     struct iov_iter *from)
 {
-	loff_t to_copy;
-	int copied;
+	int to_copy, copied;
 	u8 *to;
 
 	if (!buffer->page)
@@ -190,11 +189,8 @@ static int fill_write_buffer(struct configfs_buffer *buffer, loff_t pos,
 	if (!buffer->page)
 		return -ENOMEM;
 
-	to_copy = SIMPLE_ATTR_SIZE - 1 - pos;
-	if (to_copy <= 0)
-		return 0;
-	to = buffer->page + pos;
-	copied = copy_from_iter(to, to_copy, from);
+	to = buffer->page;
+	copied = copy_from_iter(to, SIMPLE_ATTR_SIZE - 1, from);
 	buffer->needs_read_fill = 1;
 	/* if buf is assumed to contain a string, terminate it by \0,
 	 * so e.g. sscanf() can scan the string easily */
@@ -227,14 +223,14 @@ static ssize_t configfs_write_iter(struct kiocb *iocb, struct iov_iter *from)
 {
 	struct file *file = iocb->ki_filp;
 	struct configfs_buffer *buffer = file->private_data;
-	ssize_t len;
+	int len;
 
 	mutex_lock(&buffer->mutex);
-	len = fill_write_buffer(buffer, iocb->ki_pos, from);
+	len = fill_write_buffer(buffer, from);
 	if (len > 0)
 		len = flush_write_buffer(file, buffer, len);
 	if (len > 0)
-		iocb->ki_pos += len;
+		iocb->ki_pos = len;
 	mutex_unlock(&buffer->mutex);
 	return len;
 }
#11 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Hi Bart,

I reviewed and tested the new patch. For me it works fine.

Just one warning to fix:

fs/configfs/file.c: In function ‘fill_write_buffer’:
fs/configfs/file.c:184:6: warning: unused variable ‘to_copy’ [-Wunused-variable]

  int to_copy, copied;
      ^~~~~~~

Apart from that you can add my tested-by or reviewed-by if you want.

Thank you,
Bodo

Quoted 80 lines · Bart Van Assche

Hi Bodo,

How about replacing patches 1 and 2 from this series with the patch below? Do you agree that this patch is sufficient to restore the behavior from kernel v5.13 and before?

Thanks,

Bart.

Subject: [PATCH 1/3] configfs: Restore the kernel v5.13 text attribute write behavior

Instead of writing at the offset specified by the write() system call, always write at offset zero.

Cc Bodo Stroesser <[email protected]>
Cc Martin K. Petersen <[email protected]>
Cc Yanko Kaneti <[email protected]>
Cc Brendan Higgins <[email protected]>
---
  fs/configfs/file.c | 20 ++++++++------------
  1 file changed, 8 insertions(+), 12 deletions(-)
fs/configfs/file.c +3 -4
index 5a0be9985bae..8adf6250b207 100644
@@ -177,12 +177,11 @@ static ssize_t configfs_bin_read_iter(struct kiocb 
 *iocb, struct iov_iter *to)
      return retval;
  }
 
-/* Fill [buffer, buffer + pos) with data coming from @from. */
-static int fill_write_buffer(struct configfs_buffer *buffer, loff_t pos,
+/* Fill @buffer with data coming from @from. */
+static int fill_write_buffer(struct configfs_buffer *buffer,
                   struct iov_iter *from)
  {
-    loff_t to_copy;
-    int copied;
+    int to_copy, copied;
      u8 *to;
 

     if (!buffer->page)

fs/configfs/file.c +2 -5
@@ -190,11 +189,8 @@ static int fill_write_buffer(struct configfs_buffer 
 *buffer, loff_t pos,
      if (!buffer->page)
          return -ENOMEM;
 
-    to_copy = SIMPLE_ATTR_SIZE - 1 - pos;
-    if (to_copy <= 0)
-        return 0;
-    to = buffer->page + pos;
-    copied = copy_from_iter(to, to_copy, from);
+    to = buffer->page;
+    copied = copy_from_iter(to, SIMPLE_ATTR_SIZE - 1, from);
      buffer->needs_read_fill = 1;
      /* if buf is assumed to contain a string, terminate it by \0,

      * so e.g. sscanf() can scan the string easily */

fs/configfs/file.c +3 -3
@@ -227,14 +223,14 @@ static ssize_t configfs_write_iter(struct kiocb 
 *iocb, struct iov_iter *from)
  {
      struct file *file = iocb->ki_filp;
      struct configfs_buffer *buffer = file->private_data;
-    ssize_t len;
+    int len;
 
      mutex_lock(&buffer->mutex);
-    len = fill_write_buffer(buffer, iocb->ki_pos, from);
+    len = fill_write_buffer(buffer, from);
      if (len > 0)
          len = flush_write_buffer(file, buffer, len);
      if (len > 0)
-        iocb->ki_pos += len;
+        iocb->ki_pos = len;
      mutex_unlock(&buffer->mutex);
      return len;

 }

#12 lore raw

Re: [PATCH 2/4] configfs: Fix writing at a non-zero offset

Quoted 10 lines · Bodo Stroesser

I reviewed and tested the new patch. For me it works fine.

Just one warning to fix:

fs/configfs/file.c: In function ‘fill_write_buffer’:
fs/configfs/file.c:184:6: warning: unused variable ‘to_copy’ [-Wunused-variable]
  int to_copy, copied;
      ^~~~~~~

Apart from that you can add my tested-by or reviewed-by if you want.

I will remove the 'to_copy' variable and also add your Tested-by. Thanks for having tested this patch!

Bart.

#13 lore raw

PATCH 3/4 kunit: Add support for suite initialization and cleanup

In reply to Bart Van Assche
Cc Brendan Higgins <[email protected]>
Cc Bodo Stroesser <[email protected]>
Cc Martin K. Petersen <[email protected]>
Cc Yanko Kaneti <[email protected]>
---
 include/kunit/test.h |  4 ++++
 lib/kunit/test.c     | 14 ++++++++++++++
 2 files changed, 18 insertions(+)
include/kunit/test.h +4 -0
index 24b40e5c160b..a6eef96a409c 100644
@@ -215,6 +215,8 @@ static inline char *kunit_status_to_ok_not_ok(enum kunit_status status)
  * struct kunit_suite - describes a related collection of &struct kunit_case
  *
  * @name:	the name of the test. Purely informational.
+ * @init_suite:	called once per test suite before the test cases.
+ * @exit_suite:	called once per test suite after all test cases.
  * @init:	called before every test case.
  * @exit:	called after every test case.
  * @test_cases:	a null terminated array of test cases.
@@ -229,6 +231,8 @@ static inline char *kunit_status_to_ok_not_ok(enum kunit_status status)
  */
 struct kunit_suite {
 	const char name[256];
+	int (*init_suite)(void);
+	void (*exit_suite)(void);
 	int (*init)(struct kunit *test);
 	void (*exit)(struct kunit *test);
 	struct kunit_case *test_cases;
lib/kunit/test.c +14 -0
index d79ecb86ea57..c271692ced93 100644
@@ -397,9 +397,19 @@ int kunit_run_tests(struct kunit_suite *suite)
 {
 	char param_desc[KUNIT_PARAM_DESC_SIZE];
 	struct kunit_case *test_case;
+	int res = 0;
 
 	kunit_print_subtest_start(suite);
 
+	if (suite->init_suite)
+		res = suite->init_suite();
+
+	if (res < 0) {
+		kunit_log(KERN_INFO, suite, KUNIT_SUBTEST_INDENT
+			  "# Suite initialization failed (%d)\n", res);
+		goto end;
+	}
+
 	kunit_suite_for_each_test_case(suite, test_case) {
 		struct kunit test = { .param_value = NULL, .param_index = 0 };
 		test_case->status = KUNIT_SKIPPED;
@@ -439,6 +449,10 @@ int kunit_run_tests(struct kunit_suite *suite)
 				      test.status_comment);
 	}
 
+	if (suite->exit_suite)
+		suite->exit_suite();
+
+end:
 	kunit_print_subtest_end(suite);
 
 	return 0;
Reviewed-by on v1 3/4
#14 lore raw

Re: [PATCH 3/4] kunit: Add support for suite initialization and cleanup

Quoted 5 lines · Bart Van Assche
Cc Brendan Higgins <[email protected]>
Cc Bodo Stroesser <[email protected]>
Cc Martin K. Petersen <[email protected]>
Cc Yanko Kaneti <[email protected]>

Please also CC [email protected], [email protected], [email protected], and [email protected] for KUnit changes in the future.

Quoted 1 line

This seems pretty sensible.

Quoted 19 lines
---
 include/kunit/test.h |  4 ++++
 lib/kunit/test.c     | 14 ++++++++++++++
 2 files changed, 18 insertions(+)
include/kunit/test.h +2 -0
index 24b40e5c160b..a6eef96a409c 100644
@@ -215,6 +215,8 @@ static inline char *kunit_status_to_ok_not_ok(enum kunit_status status)
  * struct kunit_suite - describes a related collection of &struct kunit_case
  *
  * @name:      the name of the test. Purely informational.
+ * @init_suite:        called once per test suite before the test cases.
+ * @exit_suite:        called once per test suite after all test cases.
  * @init:      called before every test case.
  * @exit:      called after every test case.
  * @test_cases:        a null terminated array of test cases.

include/kunit/test.h static inline char *kunit_status_to_ok_not_ok(enum kunit_status status) · line 235

Quoted 5 lines
include/kunit/test.h +2 -0
@@ -229,6 +231,8 @@ static inline char *kunit_status_to_ok_not_ok(enum kunit_status status)
  */
 struct kunit_suite {
        const char name[256];
+       int (*init_suite)(void);
+       void (*exit_suite)(void);

I like this idea. Many other unit testing libraries in other languages have something similar.

I think it probably makes sense to not use any kind of context object here (as you have done); nevertheless, I still think it is an appropriate question for the list.

Quoted 38 lines
include/kunit/test.h +0 -0
        int (*init)(struct kunit *test);
        void (*exit)(struct kunit *test);
        struct kunit_case *test_cases;
lib/kunit/test.c +14 -0
index d79ecb86ea57..c271692ced93 100644
@@ -397,9 +397,19 @@ int kunit_run_tests(struct kunit_suite *suite)
 {
        char param_desc[KUNIT_PARAM_DESC_SIZE];
        struct kunit_case *test_case;
+       int res = 0;
 
        kunit_print_subtest_start(suite);
 
+       if (suite->init_suite)
+               res = suite->init_suite();
+
+       if (res < 0) {
+               kunit_log(KERN_INFO, suite, KUNIT_SUBTEST_INDENT
+                         "# Suite initialization failed (%d)\n", res);
+               goto end;
+       }
+
        kunit_suite_for_each_test_case(suite, test_case) {
                struct kunit test = { .param_value = NULL, .param_index = 0 };
                test_case->status = KUNIT_SKIPPED;
@@ -439,6 +449,10 @@ int kunit_run_tests(struct kunit_suite *suite)
                                      test.status_comment);
        }
 
+       if (suite->exit_suite)
+               suite->exit_suite();
+
+end:
        kunit_print_subtest_end(suite);
 
        return 0;
#15 lore raw

Re: [PATCH 3/4] kunit: Add support for suite initialization and cleanup

Quoted 10 lines · Brendan Higgins
Quoted 5 lines · Bart Van Assche
Cc Brendan Higgins <[email protected]>
Cc Bodo Stroesser <[email protected]>
Cc Martin K. Petersen <[email protected]>
Cc Yanko Kaneti <[email protected]>

Please also CC [email protected], [email protected], [email protected], and [email protected] for KUnit changes in the future.

Will do.

Quoted 5 lines
Quoted 1 line

This seems pretty sensible.

Thanks for the review!

Bart.

#16 lore raw

PATCH 4/4 configfs: Add unit tests

In reply to Bart Van Assche

Testing configfs read and write behavior is non-trivial and tedious. Hence these configfs kunit tests that make it easier to test configfs text and binary attribute support. This is how I run these tests:

set -e
if [ -e .config ]; then
make ARCH=um mrproper
fi
if [ ! -e .kunit/.kunitconfig ]; then
cat <<EOF >.kunit/.kunitconfig
CONFIG_CONFIGFS_FS=y
CONFIG_CONFIGFS_KUNIT_TEST=y
CONFIG_KUNIT=y
CONFIG_PROVE_LOCKING=y
CONFIG_SYSFS=y
CONFIG_UBSAN=y
EOF
cp .kunit/.kunitconfig .kunit/.config
fi
./tools/testing/kunit/kunit.py run

Cc Bodo Stroesser <[email protected]>
Cc Martin K. Petersen <[email protected]>
Cc Yanko Kaneti <[email protected]>
Cc Brendan Higgins <[email protected]>
---
 fs/configfs/Kconfig         |   8 +
 fs/configfs/Makefile        |   2 +
 fs/configfs/configfs-test.c | 375 ++++++++++++++++++++++++++++++++++++
 3 files changed, 385 insertions(+)
 create mode 100644 fs/configfs/configfs-test.c
fs/configfs/Kconfig +8 -0
index 272b64456999..9f8f3d8ca67d 100644
@@ -10,3 +10,11 @@ config CONFIGFS_FS
 
 	  Both sysfs and configfs can and should exist together on the
 	  same system. One is not a replacement for the other.
+
+config CONFIGFS_KUNIT_TEST
+	tristate "Configfs Kunit test" if !KUNIT_ALL_TESTS
+	depends on CONFIGFS_FS && KUNIT
+	default KUNIT_ALL_TESTS
+	help
+	  Run the configfs unit tests at boot time. For more information, see
+	  also the Kunit documentation in Documentation/dev-tools/kunit/.
fs/configfs/Makefile +2 -0
index 0200498ede27..388003fa9f37 100644
@@ -6,3 +6,5 @@
 obj-$(CONFIG_CONFIGFS_FS)	+= configfs.o
 
 configfs-objs	:= inode.o file.o dir.o symlink.o mount.o item.o
+
+obj-$(CONFIG_CONFIGFS_KUNIT_TEST) += configfs-test.o
fs/configfs/configfs-test.c new file +375 -0

375 lines not on this page · raw message