[PATCH 3/7] Drivers: hv: vmbus: Fix a bug in hv_need_to_signal_on_read()

We need to issue a full memory barrier prior making a signalling decision.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Cc: stable@vger.kernel.org
---
 drivers/hv/ring_buffer.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index 2919395..67dc245 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c

@@ -104,6 +104,7 @@ static bool hv_need_to_signal_on_read(struct hv_ring_buffer_info *rbi)
 	u32 cur_write_sz;
 	u32 pending_sz;
 
+	mb();
 	pending_sz = READ_ONCE(rbi->ring_buffer->pending_send_sz);
 	/* If the other end is not blocked on write don't bother. */
 	if (pending_sz == 0)

-- 
1.7.4.1

On Wed, Mar 23, 2016 at 05:53:53PM -0700, K. Y. Srinivasan wrote:

We need to issue a full memory barrier prior making a signalling decision.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Cc: stable@vger.kernel.org
---
 drivers/hv/ring_buffer.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

why is this burried in a 7 patch series?  Don't you want this in
4.6-final?

Please break this out into a separate patch, and redo this series.

thanks,

greg k-h

On Wed, Mar 23, 2016 at 05:53:53PM -0700, K. Y. Srinivasan wrote:

And, are you sure this is correct?  You better document the heck out of
this, why it's here, and what it is protecting.  "raw" mb() calls are
really rare for good reason.

thanks,

greg k-h

-----Original Message-----
From: Greg KH [mailto:gregkh@linuxfoundation.org]
Sent: Thursday, March 31, 2016 5:00 PM
To: KY Srinivasan <kys@microsoft.com>
Cc: linux-kernel@vger.kernel.org; devel@linuxdriverproject.org;
olaf@aepfle.de; apw@canonical.com; vkuznets@redhat.com;
jasowang@redhat.com; stable@vger.kernel.org
Subject: Re: [PATCH 3/7] Drivers: hv: vmbus: Fix a bug in
hv_need_to_signal_on_read()

On Wed, Mar 23, 2016 at 05:53:53PM -0700, K. Y. Srinivasan wrote:

quoted

We need to issue a full memory barrier prior making a signalling decision.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Cc: stable@vger.kernel.org
---
 drivers/hv/ring_buffer.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

why is this burried in a 7 patch series?  Don't you want this in
4.6-final?

Please break this out into a separate patch, and redo this series.

Will do.

Thanks,

K. Y

thanks,

greg k-h

-----Original Message-----
From: Greg KH [mailto:gregkh@linuxfoundation.org]
Sent: Thursday, March 31, 2016 5:00 PM
To: KY Srinivasan <kys@microsoft.com>
Cc: linux-kernel@vger.kernel.org; devel@linuxdriverproject.org;
olaf@aepfle.de; apw@canonical.com; vkuznets@redhat.com;
jasowang@redhat.com; stable@vger.kernel.org
Subject: Re: [PATCH 3/7] Drivers: hv: vmbus: Fix a bug in
hv_need_to_signal_on_read()

On Wed, Mar 23, 2016 at 05:53:53PM -0700, K. Y. Srinivasan wrote:

quoted

We need to issue a full memory barrier prior making a signalling decision.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Cc: stable@vger.kernel.org
---
 drivers/hv/ring_buffer.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index 2919395..67dc245 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c

@@ -104,6 +104,7 @@ static bool hv_need_to_signal_on_read(struct

hv_ring_buffer_info *rbi)

quoted

 	u32 cur_write_sz;
 	u32 pending_sz;

+	mb();

And, are you sure this is correct?  You better document the heck out of
this, why it's here, and what it is protecting.  "raw" mb() calls are
really rare for good reason.

Yes; I am sure I need this barrier. This is a lockless ringbuffer code where
the reader  updates the read index while the writer only updates the
write index and they each sample the index modified by the other end to
make decisions.  Here is the reason for having this barrier:

"If the reading of the pend_sz (in the function hv_need_to_signal_on_read)
were to be reordered and read before we commit the new read index we could 
have a problem. If the host were to set the pending_sz after we have sampled pending_sz
and go to sleep before we commit the read index, we could miss sending the interrupt."

I will add the necessary documentation here.

Regards,

K. Y