[linux-azure:xenial][PATCH 0/2] [Hyper-V] Drivers: hv: vmbus: Fix ring buffer signaling

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

[linux-azure:xenial][PATCH 0/2] [Hyper-V] Drivers: hv: vmbus: Fix ring buffer signaling

Marcelo Henrique Cerri
BugLink: http://bugs.launchpad.net/bugs/1748662

I'm checking if generic kernels should receive the same fix and I will also
provide fixes for linux-azure-edge/xenial and linux-azure/bionic.

Marcelo Henrique Cerri (1):
  Revert "UBUNTU: SAUCE: vmbus: fix performance regression"

Michael Kelley (1):
  UBUNTU: SAUCE: hv: vmbus: Fix ring buffer signaling

 drivers/hv/ring_buffer.c | 42 ++++++++++++++++++++++++++----------------
 1 file changed, 26 insertions(+), 16 deletions(-)

--
2.7.4


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

[linux-azure:xenial][PATCH 1/2] Revert "UBUNTU: SAUCE: vmbus: fix performance regression"

Marcelo Henrique Cerri
BugLink: http://bugs.launchpad.net/bugs/1748662

This reverts commit 9ea3e22cb96dcca0d12213c12b503253ceb0e798.

Apply in a following commit a more comprehensive fix that was submitted
upstream instead.

Signed-off-by: Marcelo Henrique Cerri <[hidden email]>
---
 drivers/hv/ring_buffer.c | 28 ++++++++++++++--------------
 1 file changed, 14 insertions(+), 14 deletions(-)

diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index dbc76554f9f9..8214042757cb 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c
@@ -407,10 +407,10 @@ EXPORT_SYMBOL_GPL(__hv_pkt_iter_next);
 void hv_pkt_iter_close(struct vmbus_channel *channel)
 {
  struct hv_ring_buffer_info *rbi = &channel->inbound;
- u32 cached_write_sz, cur_write_sz, pending_sz;
+ u32 orig_write_sz;
 
  /* Available space before read_index update */
- cached_write_sz = hv_get_bytes_to_write(rbi);
+ orig_write_sz = hv_get_bytes_to_write(rbi);
 
  /*
  * Make sure all reads are done before we update the read index since
@@ -422,6 +422,10 @@ void hv_pkt_iter_close(struct vmbus_channel *channel)
  /* Update the position where ring buffer has been read from */
  rbi->ring_buffer->read_index = rbi->priv_read_index;
 
+ /* If more data is available then no need to signal */
+ if (hv_get_bytes_to_read(rbi))
+ return;
+
  /*
  * If the reading of the pend_sz were to be reordered and read
  * before we commit the new read index.
@@ -431,22 +435,18 @@ void hv_pkt_iter_close(struct vmbus_channel *channel)
  virt_wmb();
 
  if (rbi->ring_buffer->feature_bits.feat_pending_send_sz) {
- 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)
- return;
-
- /* If pending write will not fit, don't give false hope. */
- cur_write_sz = hv_get_bytes_to_write(rbi);
- if (cur_write_sz < pending_sz)
- return;
+ u32 pending_sz = READ_ONCE(rbi->ring_buffer->pending_send_sz);
 
  /*
  * If there was space before we began iteration, then
- * host was not blocked.
+ * host was not blocked. Also handles the case where
+ * pending_sz is zero because host has nothing pending.
  */
- if (cached_write_sz >= pending_sz)
+ if (orig_write_sz > pending_sz)
+ return;
+
+ /* If pending write will not fit, don't give false hope. */
+ if (hv_get_bytes_to_write(rbi) < pending_sz)
  return;
  }
 
--
2.7.4


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

[linux-azure:xenial][PATCH 2/2] UBUNTU: SAUCE: hv: vmbus: Fix ring buffer signaling

Marcelo Henrique Cerri
In reply to this post by Marcelo Henrique Cerri
From: Michael Kelley <[hidden email]>

BugLink: http://bugs.launchpad.net/bugs/1748662

Fix bugs in signaling the Hyper-V host when freeing space in the
host->guest ring buffer:

1. The interrupt_mask must not be used to determine whether to signal
   on the host->guest ring buffer
2. The ring buffer write_index must be read (via hv_get_bytes_to_write)
   *after* pending_send_sz is read in order to avoid a race condition
3. Comparisons with pending_send_sz must treat the "equals" case as
   not-enough-space
4. Don't signal if the pending_send_sz feature is not present. Older
   versions of Hyper-V that don't implement this feature will poll.

Fixes: 03bad714a161 ("vmbus: more host signalling avoidance")
Signed-off-by: Michael Kelley <[hidden email]>
[[hidden email]: backported to linux-azure]
Signed-off-by: Marcelo Henrique Cerri <[hidden email]>
---
 drivers/hv/ring_buffer.c | 30 ++++++++++++++++++++----------
 1 file changed, 20 insertions(+), 10 deletions(-)

diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index 8214042757cb..469d0eaa269e 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c
@@ -407,10 +407,11 @@ EXPORT_SYMBOL_GPL(__hv_pkt_iter_next);
 void hv_pkt_iter_close(struct vmbus_channel *channel)
 {
  struct hv_ring_buffer_info *rbi = &channel->inbound;
- u32 orig_write_sz;
-
- /* Available space before read_index update */
- orig_write_sz = hv_get_bytes_to_write(rbi);
+ u32 curr_write_sz;
+ u32 delta = rbi->ring_buffer->read_index < rbi->priv_read_index ?
+ (rbi->priv_read_index - rbi->ring_buffer->read_index) :
+ (rbi->ring_datasize - rbi->ring_buffer->read_index +
+ rbi->priv_read_index);
 
  /*
  * Make sure all reads are done before we update the read index since
@@ -438,18 +439,27 @@ void hv_pkt_iter_close(struct vmbus_channel *channel)
  u32 pending_sz = READ_ONCE(rbi->ring_buffer->pending_send_sz);
 
  /*
- * If there was space before we began iteration, then
- * host was not blocked. Also handles the case where
- * pending_sz is zero because host has nothing pending.
+ * Ensure the read of write_index in hv_get_bytes_to_write()
+ * happens after the read of pending_send_sz.
+ */
+ virt_rmb();
+ curr_write_sz = hv_get_bytes_to_write(rbi);
+
+ /*
+ * If there was space before we began iteration,
+ * then host was not blocked. Also handles case where
+ * pending_sz is zero then host has nothing pending
+ * and does not need to be signaled.
  */
- if (orig_write_sz > pending_sz)
+ if (curr_write_sz - delta > pending_sz)
  return;
 
  /* If pending write will not fit, don't give false hope. */
- if (hv_get_bytes_to_write(rbi) < pending_sz)
+ if (curr_write_sz <= pending_sz)
  return;
+
+ vmbus_setevent(channel);
  }
 
- vmbus_setevent(channel);
 }
 EXPORT_SYMBOL_GPL(hv_pkt_iter_close);
--
2.7.4


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

ACK/cmnt: [linux-azure:xenial][PATCH 0/2] [Hyper-V] Drivers: hv: vmbus: Fix ring buffer signaling

Stefan Bader-2
In reply to this post by Marcelo Henrique Cerri
On 14.02.2018 15:06, Marcelo Henrique Cerri wrote:

> BugLink: http://bugs.launchpad.net/bugs/1748662
>
> I'm checking if generic kernels should receive the same fix and I will also
> provide fixes for linux-azure-edge/xenial and linux-azure/bionic.
>
> Marcelo Henrique Cerri (1):
>   Revert "UBUNTU: SAUCE: vmbus: fix performance regression"
>
> Michael Kelley (1):
>   UBUNTU: SAUCE: hv: vmbus: Fix ring buffer signaling
>
>  drivers/hv/ring_buffer.c | 42 ++++++++++++++++++++++++++----------------
>  1 file changed, 26 insertions(+), 16 deletions(-)
>
Acked-by: Stefan Bader <[hidden email]>

Provided positive feedback in the bug report.


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team

signature.asc (836 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

ACK: [linux-azure:xenial][PATCH 0/2] [Hyper-V] Drivers: hv: vmbus: Fix ring buffer signaling

Thadeu Lima de Souza Cascardo-3
In reply to this post by Marcelo Henrique Cerri
Acked-by: Thadeu Lima de Souza Cascardo <[hidden email]>

--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team
Reply | Threaded
Open this post in threaded view
|

NACK: [linux-azure:xenial][PATCH 0/2] [Hyper-V] Drivers: hv: vmbus: Fix ring buffer signaling

Marcelo Henrique Cerri
In reply to this post by Marcelo Henrique Cerri
I will submit a v2 with some additional reverts as requested on the LP
bug.


--
kernel-team mailing list
[hidden email]
https://lists.ubuntu.com/mailman/listinfo/kernel-team

signature.asc (484 bytes) Download Attachment