Skip to contentSkip to conversation

soundwire: fix error handling

PATCH In review [email protected] 4 participants 9 messages Aug 29, 2020 – Sep 1, 2020 SoundWire

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 soundwire: fix error handling

From: Tom Rix <[email protected]>

clang static analysis flags this problem

stream.c:844:9: warning: Use of memory after
it is freed

        kfree(bus->defer_msg.msg->buf);
              ^~~~~~~~~~~~~~~~~~~~~~~

This happens in an error handler cleaning up memory allocated for elements in a list.

	list_for_each_entry(m_rt, &stream->master_list, stream_node) {
		bus = m_rt->bus;

		kfree(bus->defer_msg.msg->buf);
		kfree(bus->defer_msg.msg);
	}

And is triggered when the call to sdw_bank_switch() fails. There are a two problems.

First, when sdw_bank_switch() fails, though it frees memory it does not clear bus's reference 'defer_msg.msg' to that memory.

The second problem is the freeing msg->buf. In some cases msg will be NULL so this will dereference a null pointer. Need to check before freeing.

Fixes 99b8a5d608a6 ("soundwire: Add bank switch routine")
---
 drivers/soundwire/stream.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/soundwire/stream.c +5 -3
index 37290a799023..6e36deb505b1 100644
@@ -717,6 +717,7 @@ static int sdw_bank_switch(struct sdw_bus *bus, int m_rt_count)
 	kfree(wbuf);
 error_1:
 	kfree(wr_msg);
+	bus->defer_msg.msg = NULL;
 	return ret;
 }
 
@@ -840,9 +841,10 @@ static int do_bank_switch(struct sdw_stream_runtime *stream)
 error:
 	list_for_each_entry(m_rt, &stream->master_list, stream_node) {
 		bus = m_rt->bus;
-
-		kfree(bus->defer_msg.msg->buf);
-		kfree(bus->defer_msg.msg);
+		if (bus->defer_msg.msg) {
+			kfree(bus->defer_msg.msg->buf);
+			kfree(bus->defer_msg.msg);
+		}
 	}
 
 msg_unlock:
Signature · 1 line
2.18.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 41 lines · On Sat, Aug 29, 2020 at 8:35 AM <[email protected]> wrote:

From: Tom Rix <[email protected]>

clang static analysis flags this problem

stream.c:844:9: warning: Use of memory after
it is freed

        kfree(bus->defer_msg.msg->buf);
              ^~~~~~~~~~~~~~~~~~~~~~~

This happens in an error handler cleaning up memory allocated for elements in a list.

        list_for_each_entry(m_rt, &stream->master_list, stream_node) {
                bus = m_rt->bus;

                kfree(bus->defer_msg.msg->buf);
                kfree(bus->defer_msg.msg);
        }

And is triggered when the call to sdw_bank_switch() fails. There are a two problems.

First, when sdw_bank_switch() fails, though it frees memory it does not clear bus's reference 'defer_msg.msg' to that memory.

The second problem is the freeing msg->buf. In some cases msg will be NULL so this will dereference a null pointer. Need to check before freeing.

Fixes 99b8a5d608a6 ("soundwire: Add bank switch routine")
---
 drivers/soundwire/stream.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/soundwire/stream.c +0 -0
index 37290a799023..6e36deb505b1 100644

drivers/soundwire/stream.c static int sdw_bank_switch(struct sdw_bus *bus, int m_rt_count) · line 720

Quoted 4 lines
drivers/soundwire/stream.c +1 -0
@@ -717,6 +717,7 @@ static int sdw_bank_switch(struct sdw_bus *bus, int m_rt_count)
        kfree(wbuf);
 error_1:
        kfree(wr_msg);
+       bus->defer_msg.msg = NULL;

This fix looks correct to me because L668 sets bus->defer_msg.msg = wr_msg;, but on error L719 frees wr_msg, so now bus->defer_msg.msg is a dangling pointer.

Quoted 4 lines
drivers/soundwire/stream.c +0 -0
        return ret;
 }
 

drivers/soundwire/stream.c static int do_bank_switch(struct sdw_stream_runtime *stream) · line 847

Quoted 10 lines
drivers/soundwire/stream.c +4 -3
@@ -840,9 +841,10 @@ static int do_bank_switch(struct sdw_stream_runtime *stream)
 error:
        list_for_each_entry(m_rt, &stream->master_list, stream_node) {
                bus = m_rt->bus;
-
-               kfree(bus->defer_msg.msg->buf);
-               kfree(bus->defer_msg.msg);
+               if (bus->defer_msg.msg) {
+                       kfree(bus->defer_msg.msg->buf);
+                       kfree(bus->defer_msg.msg);
+               }

I'd prefer a conditional check for each, but sdw_ml_sync_bank_switch() has this same pattern, so it looks like the lifetime of these two match.

Thanks for the patch!

Quoted 6 lines
drivers/soundwire/stream.c +0 -1
        }
 
 msg_unlock:
--

2.18.1

Signature · 2 lines
Thanks,
~Nick Desaulniers
author
#3 lore raw
Quoted 8 lines · Nick Desaulniers
Quoted 4 lines · Nick Desaulniers
Quoted 3 lines · On Sat, Aug 29, 2020 at 8:35 AM <[email protected]> wrote:

From: Tom Rix <[email protected]>

clang static analysis flags this problem

Also, Tom, please use ./scripts/get_maintainer.pl on your patches to CC our mailing list; [email protected].

gcc is still doing the building, so it this appropriate ?

Asking because i have been sed-ing clang-build-linux out.

Tom

#4 lore raw
Quoted 8 lines · Tom Rix
Quoted 5 lines · Nick Desaulniers
Quoted 3 lines · Nick Desaulniers
Quoted 2 lines · On Sat, Aug 29, 2020 at 8:35 AM <[email protected]> wrote:

From: Tom Rix <[email protected]>

Quoted 7 lines
Quoted 3 lines
Quoted 1 line
Quoted 1 line

clang static analysis flags this problem

Also, Tom, please use ./scripts/get_maintainer.pl on your patches to CC our mailing list; [email protected].

gcc is still doing the building, so it this appropriate ?

Asking because i have been sed-ing clang-build-linux out.

ah, right, you can use --use-cc=clang for analyses of builds with clang. It doesn't hurt to include our mailing list, since we'd like to know if issues get reported with the analyzer itself.

Signature · 2 lines
Thanks,
~Nick Desaulniers
author
#5 lore raw
Quoted 9 lines · Nick Desaulniers
Quoted 8 lines · Tom Rix
Quoted 6 lines · Nick Desaulniers
Quoted 4 lines · Nick Desaulniers
Quoted 3 lines · On Sat, Aug 29, 2020 at 8:35 AM <[email protected]> wrote:

From: Tom Rix <[email protected]>

clang static analysis flags this problem

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Also, Tom, please use ./scripts/get_maintainer.pl on your patches to CC our mailing list; [email protected].

gcc is still doing the building, so it this appropriate ?

Asking because i have been sed-ing clang-build-linux out.

ah, right, you can use --use-cc=clang for analyses of builds with clang. It doesn't hurt to include our mailing list, since we'd like to know if issues get reported with the analyzer itself.

Ok, i'll include it.

The only real issue so far has been https://reviews.llvm.org/D83984

which fixes reporting on a couple of files.

Tom

#6 lore raw
In reply to Nick Desaulniers
Quoted 70 lines · Nick Desaulniers
Quoted 41 lines · On Sat, Aug 29, 2020 at 8:35 AM <[email protected]> wrote:

From: Tom Rix <[email protected]>

clang static analysis flags this problem

stream.c:844:9: warning: Use of memory after
it is freed

         kfree(bus->defer_msg.msg->buf);
               ^~~~~~~~~~~~~~~~~~~~~~~

This happens in an error handler cleaning up memory allocated for elements in a list.

         list_for_each_entry(m_rt, &stream->master_list, stream_node) {
                 bus = m_rt->bus;

                 kfree(bus->defer_msg.msg->buf);
                 kfree(bus->defer_msg.msg);
         }

And is triggered when the call to sdw_bank_switch() fails. There are a two problems.

First, when sdw_bank_switch() fails, though it frees memory it does not clear bus's reference 'defer_msg.msg' to that memory.

The second problem is the freeing msg->buf. In some cases msg will be NULL so this will dereference a null pointer. Need to check before freeing.

Fixes 99b8a5d608a6 ("soundwire: Add bank switch routine")
---
  drivers/soundwire/stream.c | 8 +++++---
  1 file changed, 5 insertions(+), 3 deletions(-)
drivers/soundwire/stream.c +0 -0
index 37290a799023..6e36deb505b1 100644

drivers/soundwire/stream.c static int sdw_bank_switch(struct sdw_bus *bus, int m_rt_count) · line 720

Quoted 4 lines
drivers/soundwire/stream.c +1 -0
@@ -717,6 +717,7 @@ static int sdw_bank_switch(struct sdw_bus *bus, int m_rt_count)
         kfree(wbuf);
  error_1:
         kfree(wr_msg);
+       bus->defer_msg.msg = NULL;

This fix looks correct to me because L668 sets bus->defer_msg.msg = wr_msg;, but on error L719 frees wr_msg, so now bus->defer_msg.msg is a dangling pointer.

Quoted 4 lines
drivers/soundwire/stream.c +0 -0
         return ret;
  }
 

drivers/soundwire/stream.c static int do_bank_switch(struct sdw_stream_runtime *stream) · line 847

Quoted 10 lines
drivers/soundwire/stream.c +4 -3
@@ -840,9 +841,10 @@ static int do_bank_switch(struct sdw_stream_runtime *stream)
  error:
         list_for_each_entry(m_rt, &stream->master_list, stream_node) {
                 bus = m_rt->bus;
-
-               kfree(bus->defer_msg.msg->buf);
-               kfree(bus->defer_msg.msg);
+               if (bus->defer_msg.msg) {
+                       kfree(bus->defer_msg.msg->buf);
+                       kfree(bus->defer_msg.msg);
+               }

I'd prefer a conditional check for each, but sdw_ml_sync_bank_switch() has this same pattern, so it looks like the lifetime of these two match.

Quoted 1 line

Also looks good to me.

maintainer
#7 lore raw
In reply to [email protected]

Hello Tom,

Quoted 22 lines · On 29-08-20, 08:35, [email protected] wrote:

From: Tom Rix <[email protected]>

clang static analysis flags this problem

stream.c:844:9: warning: Use of memory after
it is freed

        kfree(bus->defer_msg.msg->buf);
              ^~~~~~~~~~~~~~~~~~~~~~~

This happens in an error handler cleaning up memory allocated for elements in a list.

	list_for_each_entry(m_rt, &stream->master_list, stream_node) {
		bus = m_rt->bus;

		kfree(bus->defer_msg.msg->buf);
		kfree(bus->defer_msg.msg);
	}

And is triggered when the call to sdw_bank_switch() fails. There are a two problems.

Quoted 6 lines

First, when sdw_bank_switch() fails, though it frees memory it does not clear bus's reference 'defer_msg.msg' to that memory.

The second problem is the freeing msg->buf. In some cases msg will be NULL so this will dereference a null pointer. Need to check before freeing.

The change looks good to me, but the title of patch should be revised.

The patch subject should describe the patch, in this case is setting pointer to null on cleanup, so an appropriate subject may be" "[PATCH]: soundwire: set defer_msg to null".

Please revise subject line and update including the ack/reviews received

Thanks

Quoted 35 lines
Fixes 99b8a5d608a6 ("soundwire: Add bank switch routine")
---
 drivers/soundwire/stream.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/soundwire/stream.c +5 -3
index 37290a799023..6e36deb505b1 100644
@@ -717,6 +717,7 @@ static int sdw_bank_switch(struct sdw_bus *bus, int m_rt_count)
 	kfree(wbuf);
 error_1:
 	kfree(wr_msg);
+	bus->defer_msg.msg = NULL;
 	return ret;
 }
 
@@ -840,9 +841,10 @@ static int do_bank_switch(struct sdw_stream_runtime *stream)
 error:
 	list_for_each_entry(m_rt, &stream->master_list, stream_node) {
 		bus = m_rt->bus;
-
-		kfree(bus->defer_msg.msg->buf);
-		kfree(bus->defer_msg.msg);
+		if (bus->defer_msg.msg) {
+			kfree(bus->defer_msg.msg->buf);
+			kfree(bus->defer_msg.msg);
+		}
 	}
 
 msg_unlock:
Signature · 1 line
2.18.1
Signature · 1 line
~Vinod
#8 lore raw
Quoted 33 lines · Vinod Koul

Hello Tom,

Quoted 22 lines · On 29-08-20, 08:35, [email protected] wrote:

From: Tom Rix <[email protected]>

clang static analysis flags this problem

stream.c:844:9: warning: Use of memory after
it is freed

        kfree(bus->defer_msg.msg->buf);
              ^~~~~~~~~~~~~~~~~~~~~~~

This happens in an error handler cleaning up memory allocated for elements in a list.

      list_for_each_entry(m_rt, &stream->master_list, stream_node) {
              bus = m_rt->bus;

              kfree(bus->defer_msg.msg->buf);
              kfree(bus->defer_msg.msg);
      }

And is triggered when the call to sdw_bank_switch() fails. There are a two problems.

Quoted 6 lines

First, when sdw_bank_switch() fails, though it frees memory it does not clear bus's reference 'defer_msg.msg' to that memory.

The second problem is the freeing msg->buf. In some cases msg will be NULL so this will dereference a null pointer. Need to check before freeing.

Quoted 8 lines

The change looks good to me, but the title of patch should be revised.

The patch subject should describe the patch, in this case is setting pointer to null on cleanup, so an appropriate subject may be" "[PATCH]: soundwire: set defer_msg to null".

Please revise subject line and update including the ack/reviews received

That's fair, I think
soundwire: fix double free of dangling pointer
would be most precise.

Signature · 2 lines
Thanks,
~Nick Desaulniers