From: Dexuan Cui <[email protected]>
We can concurrently try to open the same sub-channel from 2 paths:
path #1: vmbus_onoffer() -> vmbus_process_offer() -> handle_sc_creation().
path #2: storvsc_probe() -> storvsc_connect_to_vsp() ->
-> storvsc_channel_init() -> handle_multichannel_storage() ->
-> vmbus_are_subchannels_present() -> handle_sc_creation().
They conflict with each other, but it was not an issue before the recent
commit ae6935ed7d42 ("vmbus: split ring buffer allocation from open"),
because at the beginning of vmbus_open() we checked newchannel->state so
only one path could succeed, and the other would return with -EINVAL.
After ae6935ed7d42, the failing path frees the channel's ringbuffer by
vmbus_free_ring(), and this causes a panic later.
Commit ae6935ed7d42 itself is good, and it just reveals the longstanding
race. We can resolve the issue by removing path #2, i.e. removing the
second vmbus_are_subchannels_present() in handle_multichannel_storage().
BTW, the comment "Check to see if sub-channels have already been created"
in handle_multichannel_storage() is incorrect: when we unload the driver,
we first close the sub-channel(s) and then close the primary channel, next
the host sends rescind-offer message(s) so primary->sc_list will become
empty. This means the first vmbus_are_subchannels_present() in
handle_multichannel_storage() is never useful.
Fixes: ae6935ed7d42 ("vmbus: split ring buffer allocation from open")
Cc: [email protected]
Cc: Long Li <[email protected]>
Cc: Stephen Hemminger <[email protected]>
Cc: K. Y. Srinivasan <[email protected]>
Cc: Haiyang Zhang <[email protected]>
Signed-off-by: Dexuan Cui <[email protected]>
Signed-off-by: K. Y. Srinivasan <[email protected]>
---
drivers/scsi/storvsc_drv.c | 61 +++++++++++++++++++-------------------
1 file changed, 30 insertions(+), 31 deletions(-)
diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c
index f03dc03a42c3..8f88348ebe42 100644
--- a/drivers/scsi/storvsc_drv.c
+++ b/drivers/scsi/storvsc_drv.c
@@ -446,7 +446,6 @@ struct storvsc_device {
bool destroy;
bool drain_notify;
- bool open_sub_channel;
atomic_t num_outstanding_req;
struct Scsi_Host *host;
@@ -636,33 +635,38 @@ static inline struct storvsc_device *get_in_stor_device(
static void handle_sc_creation(struct vmbus_channel *new_sc)
{
struct hv_device *device = new_sc->primary_channel->device_obj;
+ struct device *dev = &device->device;
struct storvsc_device *stor_device;
struct vmstorage_channel_properties props;
+ int ret;
stor_device = get_out_stor_device(device);
if (!stor_device)
return;
- if (stor_device->open_sub_channel == false)
- return;
-
memset(&props, 0, sizeof(struct vmstorage_channel_properties));
- vmbus_open(new_sc,
- storvsc_ringbuffer_size,
- storvsc_ringbuffer_size,
- (void *)&props,
- sizeof(struct vmstorage_channel_properties),
- storvsc_on_channel_callback, new_sc);
+ ret = vmbus_open(new_sc,
+ storvsc_ringbuffer_size,
+ storvsc_ringbuffer_size,
+ (void *)&props,
+ sizeof(struct vmstorage_channel_properties),
+ storvsc_on_channel_callback, new_sc);
- if (new_sc->state == CHANNEL_OPENED_STATE) {
- stor_device->stor_chns[new_sc->target_cpu] = new_sc;
- cpumask_set_cpu(new_sc->target_cpu, &stor_device->alloced_cpus);
+ /* In case vmbus_open() fails, we don't use the sub-channel. */
+ if (ret != 0) {
+ dev_err(dev, "Failed to open sub-channel: err=%d\n", ret);
+ return;
}
+
+ /* Add the sub-channel to the array of available channels. */
+ stor_device->stor_chns[new_sc->target_cpu] = new_sc;
+ cpumask_set_cpu(new_sc->target_cpu, &stor_device->alloced_cpus);
}
static void handle_multichannel_storage(struct hv_device *device, int max_chns)
{
+ struct device *dev = &device->device;
struct storvsc_device *stor_device;
int num_cpus = num_online_cpus();
int num_sc;
@@ -679,21 +683,11 @@ static void handle_multichannel_storage(struct hv_device *device, int max_chns)
request = &stor_device->init_request;
vstor_packet = &request->vstor_packet;
- stor_device->open_sub_channel = true;
/*
* Establish a handler for dealing with subchannels.
*/
vmbus_set_sc_create_callback(device->channel, handle_sc_creation);
- /*
- * Check to see if sub-channels have already been created. This
- * can happen when this driver is re-loaded after unloading.
- */
-
- if (vmbus_are_subchannels_present(device->channel))
- return;
-
- stor_device->open_sub_channel = false;
/*
* Request the host to create sub-channels.
*/
@@ -710,23 +704,29 @@ static void handle_multichannel_storage(struct hv_device *device, int max_chns)
VM_PKT_DATA_INBAND,
VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED);
- if (ret != 0)
+ if (ret != 0) {
+ dev_err(dev, "Failed to create sub-channel: err=%d\n", ret);
return;
+ }
t = wait_for_completion_timeout(&request->wait_event, 10*HZ);
- if (t == 0)
+ if (t == 0) {
+ dev_err(dev, "Failed to create sub-channel: timed out\n");
return;
+ }
if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
- vstor_packet->status != 0)
+ vstor_packet->status != 0) {
+ dev_err(dev, "Failed to create sub-channel: op=%d, sts=%d\n",
+ vstor_packet->operation, vstor_packet->status);
return;
+ }
/*
- * Now that we created the sub-channels, invoke the check; this
- * may trigger the callback.
+ * We need to do nothing here, because vmbus_process_offer()
+ * invokes channel->sc_creation_callback, which will open and use
+ * the sub-channel(s).
*/
- stor_device->open_sub_channel = true;
- vmbus_are_subchannels_present(device->channel);
}
static void cache_wwn(struct storvsc_device *stor_device,
@@ -1794,7 +1794,6 @@ static int storvsc_probe(struct hv_device *device,
}
stor_device->destroy = false;
- stor_device->open_sub_channel = false;
init_waitqueue_head(&stor_device->waiting_to_drain);
stor_device->device = device;
stor_device->host = host;
--
2.19.1
On Mon, Nov 26, 2018 at 12:26:17AM +0000, [email protected] wrote:
>From: Dexuan Cui <[email protected]>
>
>We can concurrently try to open the same sub-channel from 2 paths:
>
>path #1: vmbus_onoffer() -> vmbus_process_offer() -> handle_sc_creation().
>path #2: storvsc_probe() -> storvsc_connect_to_vsp() ->
> -> storvsc_channel_init() -> handle_multichannel_storage() ->
> -> vmbus_are_subchannels_present() -> handle_sc_creation().
>
>They conflict with each other, but it was not an issue before the recent
>commit ae6935ed7d42 ("vmbus: split ring buffer allocation from open"),
>because at the beginning of vmbus_open() we checked newchannel->state so
>only one path could succeed, and the other would return with -EINVAL.
>
>After ae6935ed7d42, the failing path frees the channel's ringbuffer by
>vmbus_free_ring(), and this causes a panic later.
>
>Commit ae6935ed7d42 itself is good, and it just reveals the longstanding
>race. We can resolve the issue by removing path #2, i.e. removing the
>second vmbus_are_subchannels_present() in handle_multichannel_storage().
>
>BTW, the comment "Check to see if sub-channels have already been created"
>in handle_multichannel_storage() is incorrect: when we unload the driver,
>we first close the sub-channel(s) and then close the primary channel, next
>the host sends rescind-offer message(s) so primary->sc_list will become
>empty. This means the first vmbus_are_subchannels_present() in
>handle_multichannel_storage() is never useful.
>
>Fixes: ae6935ed7d42 ("vmbus: split ring buffer allocation from open")
>Cc: [email protected]
>Cc: Long Li <[email protected]>
>Cc: Stephen Hemminger <[email protected]>
>Cc: K. Y. Srinivasan <[email protected]>
>Cc: Haiyang Zhang <[email protected]>
>Signed-off-by: Dexuan Cui <[email protected]>
>Signed-off-by: K. Y. Srinivasan <[email protected]>
Just a heads-up: ae6935ed7d42 ("vmbus: split ring buffer allocation from
open") was merged in the 4.20 merge window, so this fix won't actually
apply to any of the current stable trees.
However, it's good to have tags (fixes + cc: stable) here since this fix
might end up (for whatever reason) getting merged only for 4.21, which
will then make these tags relevant.
--
Thanks,
Sasha
KY,
> From: Dexuan Cui <[email protected]>
>
> We can concurrently try to open the same sub-channel from 2 paths:
>
> path #1: vmbus_onoffer() -> vmbus_process_offer() -> handle_sc_creation().
> path #2: storvsc_probe() -> storvsc_connect_to_vsp() ->
> -> storvsc_channel_init() -> handle_multichannel_storage() ->
> -> vmbus_are_subchannels_present() -> handle_sc_creation().
>
> They conflict with each other, but it was not an issue before the recent
> commit ae6935ed7d42 ("vmbus: split ring buffer allocation from open"),
> because at the beginning of vmbus_open() we checked newchannel->state so
> only one path could succeed, and the other would return with -EINVAL.
Applied to 4.20/scsi-fixes. Thank you!
--
Martin K. Petersen Oracle Linux Engineering