In preperation for getting rid of the osd layer; change
the code to use native wait interfaces. As part of this,
fixed the buggy implementation in the osd_wait_primitive
where the condition was cleared potentially after the
condition was signalled.
Signed-off-by: K. Y. Srinivasan <[email protected]>
Signed-off-by: Hank Janssen <[email protected]>
---
drivers/staging/hv/channel.c | 58 +++++++++-----------
drivers/staging/hv/channel_mgmt.c | 46 +++++++---------
drivers/staging/hv/channel_mgmt.h | 4 +-
drivers/staging/hv/connection.c | 28 +++++++---
drivers/staging/hv/netvsc.c | 102 ++++++++++++++++++-----------------
drivers/staging/hv/netvsc.h | 3 +-
drivers/staging/hv/rndis_filter.c | 38 +++++++++----
drivers/staging/hv/storvsc.c | 98 ++++++++++++++++++++--------------
drivers/staging/hv/vmbus_private.h | 3 +-
9 files changed, 208 insertions(+), 172 deletions(-)
diff --git a/drivers/staging/hv/channel.c b/drivers/staging/hv/channel.c
index 6c292e6..5a0923c 100644
--- a/drivers/staging/hv/channel.c
+++ b/drivers/staging/hv/channel.c
@@ -19,6 +19,8 @@
* Hank Janssen <[email protected]>
*/
#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/module.h>
@@ -243,11 +245,7 @@ int vmbus_open(struct vmbus_channel *newchannel, u32 send_ringbuffer_size,
goto errorout;
}
- openInfo->waitevent = osd_waitevent_create();
- if (!openInfo->waitevent) {
- err = -ENOMEM;
- goto errorout;
- }
+ init_waitqueue_head(&openInfo->waitevent);
openMsg = (struct vmbus_channel_open_channel *)openInfo->msg;
openMsg->header.msgtype = CHANNELMSG_OPENCHANNEL;
@@ -280,8 +278,15 @@ int vmbus_open(struct vmbus_channel *newchannel, u32 send_ringbuffer_size,
goto Cleanup;
}
- /* FIXME: Need to time-out here */
- osd_waitevent_wait(openInfo->waitevent);
+ openInfo->wait_condition = 0;
+ wait_event_timeout(openInfo->waitevent,
+ openInfo->wait_condition,
+ msecs_to_jiffies(1000));
+ if (openInfo->wait_condition == 0) {
+ err = -ETIMEDOUT;
+ goto errorout;
+ }
+
if (openInfo->response.open_result.status == 0)
DPRINT_INFO(VMBUS, "channel <%p> open success!!", newchannel);
@@ -294,7 +299,6 @@ Cleanup:
list_del(&openInfo->msglistentry);
spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
- kfree(openInfo->waitevent);
kfree(openInfo);
return 0;
@@ -509,11 +513,7 @@ int vmbus_establish_gpadl(struct vmbus_channel *channel, void *kbuffer,
if (ret)
return ret;
- msginfo->waitevent = osd_waitevent_create();
- if (!msginfo->waitevent) {
- ret = -ENOMEM;
- goto Cleanup;
- }
+ init_waitqueue_head(&msginfo->waitevent);
gpadlmsg = (struct vmbus_channel_gpadl_header *)msginfo->msg;
gpadlmsg->header.msgtype = CHANNELMSG_GPADL_HEADER;
@@ -533,6 +533,7 @@ int vmbus_establish_gpadl(struct vmbus_channel *channel, void *kbuffer,
DPRINT_DBG(VMBUS, "Sending GPADL Header - len %zd",
msginfo->msgsize - sizeof(*msginfo));
+ msginfo->wait_condition = 0;
ret = vmbus_post_msg(gpadlmsg, msginfo->msgsize -
sizeof(*msginfo));
if (ret != 0) {
@@ -566,7 +567,11 @@ int vmbus_establish_gpadl(struct vmbus_channel *channel, void *kbuffer,
}
}
- osd_waitevent_wait(msginfo->waitevent);
+ wait_event_timeout(msginfo->waitevent,
+ msginfo->wait_condition,
+ msecs_to_jiffies(1000));
+ BUG_ON(msginfo->wait_condition == 0);
+
/* At this point, we received the gpadl created msg */
DPRINT_DBG(VMBUS, "Received GPADL created "
@@ -582,7 +587,6 @@ Cleanup:
list_del(&msginfo->msglistentry);
spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
- kfree(msginfo->waitevent);
kfree(msginfo);
return ret;
}
@@ -605,11 +609,7 @@ int vmbus_teardown_gpadl(struct vmbus_channel *channel, u32 gpadl_handle)
if (!info)
return -ENOMEM;
- info->waitevent = osd_waitevent_create();
- if (!info->waitevent) {
- kfree(info);
- return -ENOMEM;
- }
+ init_waitqueue_head(&info->waitevent);
msg = (struct vmbus_channel_gpadl_teardown *)info->msg;
@@ -621,22 +621,20 @@ int vmbus_teardown_gpadl(struct vmbus_channel *channel, u32 gpadl_handle)
list_add_tail(&info->msglistentry,
&vmbus_connection.chn_msg_list);
spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
-
+ info->wait_condition = 0;
ret = vmbus_post_msg(msg,
sizeof(struct vmbus_channel_gpadl_teardown));
- if (ret != 0) {
- /* TODO: */
- /* something... */
- }
- osd_waitevent_wait(info->waitevent);
+ BUG_ON(ret != 0);
+ wait_event_timeout(info->waitevent,
+ info->wait_condition, msecs_to_jiffies(1000));
+ BUG_ON(info->wait_condition == 0);
/* Received a torndown response */
spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
list_del(&info->msglistentry);
spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
- kfree(info->waitevent);
kfree(info);
return ret;
}
@@ -664,18 +662,14 @@ void vmbus_close(struct vmbus_channel *channel)
if (!info)
return;
- /* info->waitEvent = osd_waitevent_create(); */
msg = (struct vmbus_channel_close_channel *)info->msg;
msg->header.msgtype = CHANNELMSG_CLOSECHANNEL;
msg->child_relid = channel->offermsg.child_relid;
ret = vmbus_post_msg(msg, sizeof(struct vmbus_channel_close_channel));
- if (ret != 0) {
- /* TODO: */
- /* something... */
- }
+ BUG_ON(ret != 0);
/* Tear down the gpadl for the channel's ring buffer */
if (channel->ringbuffer_gpadlhandle)
vmbus_teardown_gpadl(channel,
diff --git a/drivers/staging/hv/channel_mgmt.c b/drivers/staging/hv/channel_mgmt.c
index a9c9d49..da1e56a 100644
--- a/drivers/staging/hv/channel_mgmt.c
+++ b/drivers/staging/hv/channel_mgmt.c
@@ -19,6 +19,8 @@
* Hank Janssen <[email protected]>
*/
#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/list.h>
@@ -593,7 +595,8 @@ static void vmbus_onopen_result(struct vmbus_channel_message_header *hdr)
memcpy(&msginfo->response.open_result,
result,
sizeof(struct vmbus_channel_open_result));
- osd_waitevent_set(msginfo->waitevent);
+ msginfo->wait_condition = 1;
+ wake_up(&msginfo->waitevent);
break;
}
}
@@ -643,7 +646,8 @@ static void vmbus_ongpadl_created(struct vmbus_channel_message_header *hdr)
memcpy(&msginfo->response.gpadl_created,
gpadlcreated,
sizeof(struct vmbus_channel_gpadl_created));
- osd_waitevent_set(msginfo->waitevent);
+ msginfo->wait_condition = 1;
+ wake_up(&msginfo->waitevent);
break;
}
}
@@ -689,7 +693,8 @@ static void vmbus_ongpadl_torndown(
memcpy(&msginfo->response.gpadl_torndown,
gpadl_torndown,
sizeof(struct vmbus_channel_gpadl_torndown));
- osd_waitevent_set(msginfo->waitevent);
+ msginfo->wait_condition = 1;
+ wake_up(&msginfo->waitevent);
break;
}
}
@@ -730,7 +735,8 @@ static void vmbus_onversion_response(
memcpy(&msginfo->response.version_response,
version_response,
sizeof(struct vmbus_channel_version_response));
- osd_waitevent_set(msginfo->waitevent);
+ msginfo->wait_condition = 1;
+ wake_up(&msginfo->waitevent);
}
}
spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
@@ -805,44 +811,34 @@ int vmbus_request_offers(void)
if (!msginfo)
return -ENOMEM;
- msginfo->waitevent = osd_waitevent_create();
- if (!msginfo->waitevent) {
- kfree(msginfo);
- return -ENOMEM;
- }
+ init_waitqueue_head(&msginfo->waitevent);
msg = (struct vmbus_channel_message_header *)msginfo->msg;
msg->msgtype = CHANNELMSG_REQUESTOFFERS;
- /*SpinlockAcquire(gVmbusConnection.channelMsgLock);
- INSERT_TAIL_LIST(&gVmbusConnection.channelMsgList,
- &msgInfo->msgListEntry);
- SpinlockRelease(gVmbusConnection.channelMsgLock);*/
ret = vmbus_post_msg(msg,
sizeof(struct vmbus_channel_message_header));
if (ret != 0) {
DPRINT_ERR(VMBUS, "Unable to request offers - %d", ret);
- /*SpinlockAcquire(gVmbusConnection.channelMsgLock);
- REMOVE_ENTRY_LIST(&msgInfo->msgListEntry);
- SpinlockRelease(gVmbusConnection.channelMsgLock);*/
+ goto cleanup;
+ }
- goto Cleanup;
+ msginfo->wait_condition = 0;
+ wait_event_timeout(msginfo->waitevent, msginfo->wait_condition,
+ msecs_to_jiffies(1000));
+ if (msginfo->wait_condition == 0) {
+ ret = -ETIMEDOUT;
+ goto cleanup;
}
- /* osd_waitevent_wait(msgInfo->waitEvent); */
- /*SpinlockAcquire(gVmbusConnection.channelMsgLock);
- REMOVE_ENTRY_LIST(&msgInfo->msgListEntry);
- SpinlockRelease(gVmbusConnection.channelMsgLock);*/
-Cleanup:
- if (msginfo) {
- kfree(msginfo->waitevent);
+cleanup:
+ if (msginfo)
kfree(msginfo);
- }
return ret;
}
diff --git a/drivers/staging/hv/channel_mgmt.h b/drivers/staging/hv/channel_mgmt.h
index fe40bf2..3368bf1 100644
--- a/drivers/staging/hv/channel_mgmt.h
+++ b/drivers/staging/hv/channel_mgmt.h
@@ -289,8 +289,8 @@ struct vmbus_channel_msginfo {
struct list_head submsglist;
/* Synchronize the request/response if needed */
- struct osd_waitevent *waitevent;
-
+ int wait_condition;
+ wait_queue_head_t waitevent;
union {
struct vmbus_channel_version_supported version_supported;
struct vmbus_channel_open_result open_result;
diff --git a/drivers/staging/hv/connection.c b/drivers/staging/hv/connection.c
index ed0976a..51dd362 100644
--- a/drivers/staging/hv/connection.c
+++ b/drivers/staging/hv/connection.c
@@ -21,6 +21,8 @@
*
*/
#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>
@@ -97,11 +99,7 @@ int vmbus_connect(void)
goto Cleanup;
}
- msginfo->waitevent = osd_waitevent_create();
- if (!msginfo->waitevent) {
- ret = -ENOMEM;
- goto Cleanup;
- }
+ init_waitqueue_head(&msginfo->waitevent);
msg = (struct vmbus_channel_initiate_contact *)msginfo->msg;
@@ -131,14 +129,30 @@ int vmbus_connect(void)
ret = vmbus_post_msg(msg,
sizeof(struct vmbus_channel_initiate_contact));
if (ret != 0) {
+ spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
list_del(&msginfo->msglistentry);
+ spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock,
+ flags);
goto Cleanup;
}
/* Wait for the connection response */
- osd_waitevent_wait(msginfo->waitevent);
+ msginfo->wait_condition = 0;
+ wait_event_timeout(msginfo->waitevent, msginfo->wait_condition,
+ msecs_to_jiffies(1000));
+ if (msginfo->wait_condition == 0) {
+ spin_lock_irqsave(&vmbus_connection.channelmsg_lock,
+ flags);
+ list_del(&msginfo->msglistentry);
+ spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock,
+ flags);
+ ret = -ETIMEDOUT;
+ goto Cleanup;
+ }
+ spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
list_del(&msginfo->msglistentry);
+ spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
/* Check if successful */
if (msginfo->response.version_response.version_supported) {
@@ -153,7 +167,6 @@ int vmbus_connect(void)
goto Cleanup;
}
- kfree(msginfo->waitevent);
kfree(msginfo);
return 0;
@@ -174,7 +187,6 @@ Cleanup:
}
if (msginfo) {
- kfree(msginfo->waitevent);
kfree(msginfo);
}
diff --git a/drivers/staging/hv/netvsc.c b/drivers/staging/hv/netvsc.c
index a271aa7..7233564 100644
--- a/drivers/staging/hv/netvsc.c
+++ b/drivers/staging/hv/netvsc.c
@@ -19,6 +19,8 @@
* Hank Janssen <[email protected]>
*/
#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
#include <linux/mm.h>
#include <linux/delay.h>
#include <linux/io.h>
@@ -230,7 +232,7 @@ static int netvsc_init_recv_buf(struct hv_device *device)
"unable to allocate receive buffer of size %d",
net_device->recv_buf_size);
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
/* page-aligned buffer */
/* ASSERT(((unsigned long)netDevice->ReceiveBuffer & (PAGE_SIZE - 1)) == */
@@ -249,10 +251,9 @@ static int netvsc_init_recv_buf(struct hv_device *device)
if (ret != 0) {
DPRINT_ERR(NETVSC,
"unable to establish receive buffer's gpadl");
- goto Cleanup;
+ goto cleanup;
}
- /* osd_waitevent_wait(ext->ChannelInitEvent); */
/* Notify the NetVsp of the gpadl handle */
DPRINT_INFO(NETVSC, "Sending NvspMessage1TypeSendReceiveBuffer...");
@@ -268,6 +269,7 @@ static int netvsc_init_recv_buf(struct hv_device *device)
send_recv_buf.id = NETVSC_RECEIVE_BUFFER_ID;
/* Send the gpadl notification request */
+ net_device->wait_condition = 0;
ret = vmbus_sendpacket(device->channel, init_packet,
sizeof(struct nvsp_message),
(unsigned long)init_packet,
@@ -276,10 +278,14 @@ static int netvsc_init_recv_buf(struct hv_device *device)
if (ret != 0) {
DPRINT_ERR(NETVSC,
"unable to send receive buffer's gpadl to netvsp");
- goto Cleanup;
+ goto cleanup;
}
- osd_waitevent_wait(net_device->channel_init_event);
+ wait_event_timeout(net_device->channel_init_wait,
+ net_device->wait_condition,
+ msecs_to_jiffies(1000));
+ BUG_ON(net_device->wait_condition == 0);
+
/* Check the response */
if (init_packet->msg.v1_msg.
@@ -289,7 +295,7 @@ static int netvsc_init_recv_buf(struct hv_device *device)
init_packet->msg.v1_msg.
send_recv_buf_complete.status);
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
/* Parse the response */
@@ -303,7 +309,7 @@ static int netvsc_init_recv_buf(struct hv_device *device)
* sizeof(struct nvsp_1_receive_buffer_section), GFP_KERNEL);
if (net_device->recv_section == NULL) {
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
memcpy(net_device->recv_section,
@@ -327,15 +333,15 @@ static int netvsc_init_recv_buf(struct hv_device *device)
if (net_device->recv_section_cnt != 1 ||
net_device->recv_section->offset != 0) {
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
- goto Exit;
+ goto exit;
-Cleanup:
+cleanup:
netvsc_destroy_recv_buf(net_device);
-Exit:
+exit:
put_net_device(device);
return ret;
}
@@ -354,7 +360,7 @@ static int netvsc_init_send_buf(struct hv_device *device)
}
if (net_device->send_buf_size <= 0) {
ret = -EINVAL;
- goto Cleanup;
+ goto cleanup;
}
/* page-size grandularity */
@@ -367,7 +373,7 @@ static int netvsc_init_send_buf(struct hv_device *device)
DPRINT_ERR(NETVSC, "unable to allocate send buffer of size %d",
net_device->send_buf_size);
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
/* page-aligned buffer */
/* ASSERT(((unsigned long)netDevice->SendBuffer & (PAGE_SIZE - 1)) == 0); */
@@ -384,11 +390,9 @@ static int netvsc_init_send_buf(struct hv_device *device)
&net_device->send_buf_gpadl_handle);
if (ret != 0) {
DPRINT_ERR(NETVSC, "unable to establish send buffer's gpadl");
- goto Cleanup;
+ goto cleanup;
}
- /* osd_waitevent_wait(ext->ChannelInitEvent); */
-
/* Notify the NetVsp of the gpadl handle */
DPRINT_INFO(NETVSC, "Sending NvspMessage1TypeSendSendBuffer...");
@@ -403,6 +407,7 @@ static int netvsc_init_send_buf(struct hv_device *device)
NETVSC_SEND_BUFFER_ID;
/* Send the gpadl notification request */
+ net_device->wait_condition = 0;
ret = vmbus_sendpacket(device->channel, init_packet,
sizeof(struct nvsp_message),
(unsigned long)init_packet,
@@ -411,10 +416,13 @@ static int netvsc_init_send_buf(struct hv_device *device)
if (ret != 0) {
DPRINT_ERR(NETVSC,
"unable to send receive buffer's gpadl to netvsp");
- goto Cleanup;
+ goto cleanup;
}
- osd_waitevent_wait(net_device->channel_init_event);
+ wait_event_timeout(net_device->channel_init_wait,
+ net_device->wait_condition,
+ msecs_to_jiffies(1000));
+ BUG_ON(net_device->wait_condition == 0);
/* Check the response */
if (init_packet->msg.v1_msg.
@@ -424,18 +432,18 @@ static int netvsc_init_send_buf(struct hv_device *device)
init_packet->msg.v1_msg.
send_send_buf_complete.status);
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
net_device->send_section_size = init_packet->
msg.v1_msg.send_send_buf_complete.section_size;
- goto Exit;
+ goto exit;
-Cleanup:
+cleanup:
netvsc_destroy_send_buf(net_device);
-Exit:
+exit:
put_net_device(device);
return ret;
}
@@ -611,6 +619,7 @@ static int netvsc_connect_vsp(struct hv_device *device)
DPRINT_INFO(NETVSC, "Sending NvspMessageTypeInit...");
/* Send the init request */
+ net_device->wait_condition = 0;
ret = vmbus_sendpacket(device->channel, init_packet,
sizeof(struct nvsp_message),
(unsigned long)init_packet,
@@ -619,10 +628,16 @@ static int netvsc_connect_vsp(struct hv_device *device)
if (ret != 0) {
DPRINT_ERR(NETVSC, "unable to send NvspMessageTypeInit");
- goto Cleanup;
+ goto cleanup;
}
- osd_waitevent_wait(net_device->channel_init_event);
+ wait_event_timeout(net_device->channel_init_wait,
+ net_device->wait_condition,
+ msecs_to_jiffies(1000));
+ if (net_device->wait_condition == 0) {
+ ret = -ETIMEDOUT;
+ goto cleanup;
+ }
/* Now, check the response */
/* ASSERT(initPacket->Messages.InitMessages.InitComplete.MaximumMdlChainLength <= MAX_MULTIPAGE_BUFFER_COUNT); */
@@ -637,7 +652,7 @@ static int netvsc_connect_vsp(struct hv_device *device)
"unable to initialize with netvsp (status 0x%x)",
init_packet->msg.init_msg.init_complete.status);
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
if (init_packet->msg.init_msg.init_complete.
@@ -647,7 +662,7 @@ static int netvsc_connect_vsp(struct hv_device *device)
init_packet->msg.init_msg.
init_complete.negotiated_protocol_ver);
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
DPRINT_INFO(NETVSC, "Sending NvspMessage1TypeSendNdisVersion...");
@@ -666,29 +681,22 @@ static int netvsc_connect_vsp(struct hv_device *device)
/* Send the init request */
ret = vmbus_sendpacket(device->channel, init_packet,
- sizeof(struct nvsp_message),
- (unsigned long)init_packet,
- VM_PKT_DATA_INBAND, 0);
+ sizeof(struct nvsp_message),
+ (unsigned long)init_packet,
+ VM_PKT_DATA_INBAND, 0);
if (ret != 0) {
DPRINT_ERR(NETVSC,
"unable to send NvspMessage1TypeSendNdisVersion");
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
- /*
- * BUGBUG - We have to wait for the above msg since the
- * netvsp uses KMCL which acknowledges packet (completion
- * packet) since our Vmbus always set the
- * VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED flag
- */
- /* osd_waitevent_wait(NetVscChannel->ChannelInitEvent); */
/* Post the big receive buffer to NetVSP */
ret = netvsc_init_recv_buf(device);
if (ret == 0)
ret = netvsc_init_send_buf(device);
-Cleanup:
+cleanup:
put_net_device(device);
return ret;
}
@@ -715,7 +723,7 @@ static int netvsc_device_add(struct hv_device *device, void *additional_info)
net_device = alloc_net_device(device);
if (!net_device) {
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
DPRINT_DBG(NETVSC, "netvsc channel object allocated - %p", net_device);
@@ -741,11 +749,7 @@ static int netvsc_device_add(struct hv_device *device, void *additional_info)
list_add_tail(&packet->list_ent,
&net_device->recv_pkt_list);
}
- net_device->channel_init_event = osd_waitevent_create();
- if (!net_device->channel_init_event) {
- ret = -ENOMEM;
- goto Cleanup;
- }
+ init_waitqueue_head(&net_device->channel_init_wait);
/* Open the channel */
ret = vmbus_open(device->channel, net_driver->ring_buf_size,
@@ -755,7 +759,7 @@ static int netvsc_device_add(struct hv_device *device, void *additional_info)
if (ret != 0) {
DPRINT_ERR(NETVSC, "unable to open channel: %d", ret);
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
/* Channel is opened */
@@ -778,11 +782,9 @@ close:
/* Now, we can close the channel safely */
vmbus_close(device->channel);
-Cleanup:
+cleanup:
if (net_device) {
- kfree(net_device->channel_init_event);
-
list_for_each_entry_safe(packet, pos,
&net_device->recv_pkt_list,
list_ent) {
@@ -847,7 +849,6 @@ static int netvsc_device_remove(struct hv_device *device)
kfree(netvsc_packet);
}
- kfree(net_device->channel_init_event);
free_net_device(net_device);
return 0;
}
@@ -887,7 +888,8 @@ static void netvsc_send_completion(struct hv_device *device,
/* Copy the response back */
memcpy(&net_device->channel_init_pkt, nvsp_packet,
sizeof(struct nvsp_message));
- osd_waitevent_set(net_device->channel_init_event);
+ net_device->wait_condition = 1;
+ wake_up(&net_device->channel_init_wait);
} else if (nvsp_packet->hdr.msg_type ==
NVSP_MSG1_TYPE_SEND_RNDIS_PKT_COMPLETE) {
/* Get the send context */
diff --git a/drivers/staging/hv/netvsc.h b/drivers/staging/hv/netvsc.h
index 5f6dcf1..45d24b9 100644
--- a/drivers/staging/hv/netvsc.h
+++ b/drivers/staging/hv/netvsc.h
@@ -318,7 +318,8 @@ struct netvsc_device {
struct nvsp_1_receive_buffer_section *recv_section;
/* Used for NetVSP initialization protocol */
- struct osd_waitevent *channel_init_event;
+ int wait_condition;
+ wait_queue_head_t channel_init_wait;
struct nvsp_message channel_init_pkt;
struct nvsp_message revoke_packet;
diff --git a/drivers/staging/hv/rndis_filter.c b/drivers/staging/hv/rndis_filter.c
index 287e12e..e3bf004 100644
--- a/drivers/staging/hv/rndis_filter.c
+++ b/drivers/staging/hv/rndis_filter.c
@@ -19,6 +19,8 @@
* Hank Janssen <[email protected]>
*/
#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
#include <linux/highmem.h>
#include <linux/slab.h>
#include <linux/io.h>
@@ -57,7 +59,8 @@ struct rndis_device {
struct rndis_request {
struct list_head list_ent;
- struct osd_waitevent *waitevent;
+ int wait_condition;
+ wait_queue_head_t wait_event;
/*
* FIXME: We assumed a fixed size response here. If we do ever need to
@@ -129,11 +132,7 @@ static struct rndis_request *get_rndis_request(struct rndis_device *dev,
if (!request)
return NULL;
- request->waitevent = osd_waitevent_create();
- if (!request->waitevent) {
- kfree(request);
- return NULL;
- }
+ init_waitqueue_head(&request->wait_event);
rndis_msg = &request->request_msg;
rndis_msg->ndis_msg_type = msg_type;
@@ -164,7 +163,6 @@ static void put_rndis_request(struct rndis_device *dev,
list_del(&req->list_ent);
spin_unlock_irqrestore(&dev->request_lock, flags);
- kfree(req->waitevent);
kfree(req);
}
@@ -321,7 +319,8 @@ static void rndis_filter_receive_response(struct rndis_device *dev,
}
}
- osd_waitevent_set(request->waitevent);
+ request->wait_condition = 1;
+ wake_up(&request->wait_event);
} else {
DPRINT_ERR(NETVSC, "no rndis request found for this response "
"(id 0x%x res type 0x%x)",
@@ -503,11 +502,17 @@ static int rndis_filter_query_device(struct rndis_device *dev, u32 oid,
query->info_buflen = 0;
query->dev_vc_handle = 0;
+ request->wait_condition = 0;
ret = rndis_filter_send_request(dev, request);
if (ret != 0)
goto Cleanup;
- osd_waitevent_wait(request->waitevent);
+ wait_event_timeout(request->wait_event, request->wait_condition,
+ msecs_to_jiffies(1000));
+ if (request->wait_condition == 0) {
+ ret = -ETIMEDOUT;
+ goto Cleanup;
+ }
/* Copy the response back */
query_complete = &request->response_msg.msg.query_complete;
@@ -578,12 +583,14 @@ static int rndis_filter_set_packet_filter(struct rndis_device *dev,
memcpy((void *)(unsigned long)set + sizeof(struct rndis_set_request),
&new_filter, sizeof(u32));
+ request->wait_condition = 0;
ret = rndis_filter_send_request(dev, request);
if (ret != 0)
goto Cleanup;
- ret = osd_waitevent_waitex(request->waitevent, 2000/*2sec*/);
- if (!ret) {
+ wait_event_timeout(request->wait_event, request->wait_condition,
+ msecs_to_jiffies(2000));
+ if (request->wait_condition == 0) {
ret = -1;
DPRINT_ERR(NETVSC, "timeout before we got a set response...");
/*
@@ -669,13 +676,20 @@ static int rndis_filter_init_device(struct rndis_device *dev)
dev->state = RNDIS_DEV_INITIALIZING;
+ request->wait_condition = 0;
ret = rndis_filter_send_request(dev, request);
if (ret != 0) {
dev->state = RNDIS_DEV_UNINITIALIZED;
goto Cleanup;
}
- osd_waitevent_wait(request->waitevent);
+
+ wait_event_timeout(request->wait_event, request->wait_condition,
+ msecs_to_jiffies(1000));
+ if (request->wait_condition == 0) {
+ ret = -ETIMEDOUT;
+ goto Cleanup;
+ }
init_complete = &request->response_msg.msg.init_complete;
status = init_complete->status;
diff --git a/drivers/staging/hv/storvsc.c b/drivers/staging/hv/storvsc.c
index a612109..2560342 100644
--- a/drivers/staging/hv/storvsc.c
+++ b/drivers/staging/hv/storvsc.c
@@ -19,6 +19,8 @@
* Hank Janssen <[email protected]>
*/
#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/mm.h>
@@ -38,7 +40,8 @@ struct storvsc_request_extension {
struct hv_device *device;
/* Synchronize the request/response if needed */
- struct osd_waitevent *wait_event;
+ int wait_condition;
+ wait_queue_head_t wait_event;
struct vstor_packet vstor_packet;
};
@@ -200,21 +203,13 @@ static int stor_vsc_channel_init(struct hv_device *device)
* channel
*/
memset(request, 0, sizeof(struct storvsc_request_extension));
- request->wait_event = osd_waitevent_create();
- if (!request->wait_event) {
- ret = -ENOMEM;
- goto nomem;
- }
-
+ init_waitqueue_head(&request->wait_event);
vstor_packet->operation = VSTOR_OPERATION_BEGIN_INITIALIZATION;
vstor_packet->flags = REQUEST_COMPLETION_FLAG;
- /*SpinlockAcquire(gDriverExt.packetListLock);
- INSERT_TAIL_LIST(&gDriverExt.packetList, &packet->listEntry.entry);
- SpinlockRelease(gDriverExt.packetListLock);*/
-
DPRINT_INFO(STORVSC, "BEGIN_INITIALIZATION_OPERATION...");
+ request->wait_condition = 0;
ret = vmbus_sendpacket(device->channel, vstor_packet,
sizeof(struct vstor_packet),
(unsigned long)request,
@@ -223,17 +218,23 @@ static int stor_vsc_channel_init(struct hv_device *device)
if (ret != 0) {
DPRINT_ERR(STORVSC,
"unable to send BEGIN_INITIALIZATION_OPERATION");
- goto Cleanup;
+ goto cleanup;
+ }
+
+ wait_event_timeout(request->wait_event, request->wait_condition,
+ msecs_to_jiffies(1000));
+ if (request->wait_condition == 0) {
+ ret = -ETIMEDOUT;
+ goto cleanup;
}
- osd_waitevent_wait(request->wait_event);
if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
vstor_packet->status != 0) {
DPRINT_ERR(STORVSC, "BEGIN_INITIALIZATION_OPERATION failed "
"(op %d status 0x%x)",
vstor_packet->operation, vstor_packet->status);
- goto Cleanup;
+ goto cleanup;
}
DPRINT_INFO(STORVSC, "QUERY_PROTOCOL_VERSION_OPERATION...");
@@ -246,6 +247,7 @@ static int stor_vsc_channel_init(struct hv_device *device)
vstor_packet->version.major_minor = VMSTOR_PROTOCOL_VERSION_CURRENT;
FILL_VMSTOR_REVISION(vstor_packet->version.revision);
+ request->wait_condition = 0;
ret = vmbus_sendpacket(device->channel, vstor_packet,
sizeof(struct vstor_packet),
(unsigned long)request,
@@ -254,10 +256,15 @@ static int stor_vsc_channel_init(struct hv_device *device)
if (ret != 0) {
DPRINT_ERR(STORVSC,
"unable to send BEGIN_INITIALIZATION_OPERATION");
- goto Cleanup;
+ goto cleanup;
}
- osd_waitevent_wait(request->wait_event);
+ wait_event_timeout(request->wait_event, request->wait_condition,
+ msecs_to_jiffies(1000));
+ if (request->wait_condition == 0) {
+ ret = -ETIMEDOUT;
+ goto cleanup;
+ }
/* TODO: Check returned version */
if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
@@ -265,7 +272,7 @@ static int stor_vsc_channel_init(struct hv_device *device)
DPRINT_ERR(STORVSC, "QUERY_PROTOCOL_VERSION_OPERATION failed "
"(op %d status 0x%x)",
vstor_packet->operation, vstor_packet->status);
- goto Cleanup;
+ goto cleanup;
}
/* Query channel properties */
@@ -277,6 +284,7 @@ static int stor_vsc_channel_init(struct hv_device *device)
vstor_packet->storage_channel_properties.port_number =
stor_device->port_number;
+ request->wait_condition = 0;
ret = vmbus_sendpacket(device->channel, vstor_packet,
sizeof(struct vstor_packet),
(unsigned long)request,
@@ -286,10 +294,15 @@ static int stor_vsc_channel_init(struct hv_device *device)
if (ret != 0) {
DPRINT_ERR(STORVSC,
"unable to send QUERY_PROPERTIES_OPERATION");
- goto Cleanup;
+ goto cleanup;
}
- osd_waitevent_wait(request->wait_event);
+ wait_event_timeout(request->wait_event, request->wait_condition,
+ msecs_to_jiffies(1000));
+ if (request->wait_condition == 0) {
+ ret = -ETIMEDOUT;
+ goto cleanup;
+ }
/* TODO: Check returned version */
if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
@@ -297,7 +310,7 @@ static int stor_vsc_channel_init(struct hv_device *device)
DPRINT_ERR(STORVSC, "QUERY_PROPERTIES_OPERATION failed "
"(op %d status 0x%x)",
vstor_packet->operation, vstor_packet->status);
- goto Cleanup;
+ goto cleanup;
}
stor_device->path_id = vstor_packet->storage_channel_properties.path_id;
@@ -314,6 +327,7 @@ static int stor_vsc_channel_init(struct hv_device *device)
vstor_packet->operation = VSTOR_OPERATION_END_INITIALIZATION;
vstor_packet->flags = REQUEST_COMPLETION_FLAG;
+ request->wait_condition = 0;
ret = vmbus_sendpacket(device->channel, vstor_packet,
sizeof(struct vstor_packet),
(unsigned long)request,
@@ -323,25 +337,27 @@ static int stor_vsc_channel_init(struct hv_device *device)
if (ret != 0) {
DPRINT_ERR(STORVSC,
"unable to send END_INITIALIZATION_OPERATION");
- goto Cleanup;
+ goto cleanup;
}
- osd_waitevent_wait(request->wait_event);
+ wait_event_timeout(request->wait_event, request->wait_condition,
+ msecs_to_jiffies(1000));
+ if (request->wait_condition == 0) {
+ ret = -ETIMEDOUT;
+ goto cleanup;
+ }
if (vstor_packet->operation != VSTOR_OPERATION_COMPLETE_IO ||
vstor_packet->status != 0) {
DPRINT_ERR(STORVSC, "END_INITIALIZATION_OPERATION failed "
"(op %d status 0x%x)",
vstor_packet->operation, vstor_packet->status);
- goto Cleanup;
+ goto cleanup;
}
DPRINT_INFO(STORVSC, "**** storage channel up and running!! ****");
-Cleanup:
- kfree(request->wait_event);
- request->wait_event = NULL;
-nomem:
+cleanup:
put_stor_device(device);
return ret;
}
@@ -476,8 +492,8 @@ static void stor_vsc_on_channel_callback(void *context)
memcpy(&request->vstor_packet, packet,
sizeof(struct vstor_packet));
-
- osd_waitevent_set(request->wait_event);
+ request->wait_condition = 1;
+ wake_up(&request->wait_event);
} else {
stor_vsc_on_receive(device,
(struct vstor_packet *)packet,
@@ -539,7 +555,7 @@ static int stor_vsc_on_device_add(struct hv_device *device,
stor_device = alloc_stor_device(device);
if (!stor_device) {
ret = -1;
- goto Cleanup;
+ goto cleanup;
}
/* Save the channel properties to our storvsc channel */
@@ -569,7 +585,7 @@ static int stor_vsc_on_device_add(struct hv_device *device,
stor_device->port_number, stor_device->path_id,
stor_device->target_id);
-Cleanup:
+cleanup:
return ret;
}
@@ -629,16 +645,13 @@ int stor_vsc_on_host_reset(struct hv_device *device)
request = &stor_device->reset_request;
vstor_packet = &request->vstor_packet;
- request->wait_event = osd_waitevent_create();
- if (!request->wait_event) {
- ret = -ENOMEM;
- goto Cleanup;
- }
+ init_waitqueue_head(&request->wait_event);
vstor_packet->operation = VSTOR_OPERATION_RESET_BUS;
vstor_packet->flags = REQUEST_COMPLETION_FLAG;
vstor_packet->vm_srb.path_id = stor_device->path_id;
+ request->wait_condition = 0;
ret = vmbus_sendpacket(device->channel, vstor_packet,
sizeof(struct vstor_packet),
(unsigned long)&stor_device->reset_request,
@@ -647,13 +660,16 @@ int stor_vsc_on_host_reset(struct hv_device *device)
if (ret != 0) {
DPRINT_ERR(STORVSC, "Unable to send reset packet %p ret %d",
vstor_packet, ret);
- goto Cleanup;
+ goto cleanup;
}
- /* FIXME: Add a timeout */
- osd_waitevent_wait(request->wait_event);
+ wait_event_timeout(request->wait_event, request->wait_condition,
+ msecs_to_jiffies(1000));
+ if (request->wait_condition == 0) {
+ ret = -ETIMEDOUT;
+ goto cleanup;
+ }
- kfree(request->wait_event);
DPRINT_INFO(STORVSC, "host adapter reset completed");
/*
@@ -661,7 +677,7 @@ int stor_vsc_on_host_reset(struct hv_device *device)
* should have been flushed out and return to us
*/
-Cleanup:
+cleanup:
put_stor_device(device);
return ret;
}
diff --git a/drivers/staging/hv/vmbus_private.h b/drivers/staging/hv/vmbus_private.h
index 004d8de..9f505c4 100644
--- a/drivers/staging/hv/vmbus_private.h
+++ b/drivers/staging/hv/vmbus_private.h
@@ -91,7 +91,8 @@ struct vmbus_msginfo {
struct list_head msglist_entry;
/* Synchronize the request/response if needed */
- struct osd_waitevent *wait_event;
+ int wait_condition;
+ wait_queue_head_t wait_event;
/* The message itself */
unsigned char msg[0];
--
1.5.5.6
On 02/11/2011 06:59 PM, K. Y. Srinivasan wrote:
> In preperation for getting rid of the osd layer; change
> the code to use native wait interfaces. As part of this,
> fixed the buggy implementation in the osd_wait_primitive
> where the condition was cleared potentially after the
> condition was signalled.
...
> @@ -566,7 +567,11 @@ int vmbus_establish_gpadl(struct vmbus_channel *channel, void *kbuffer,
>
> }
> }
> - osd_waitevent_wait(msginfo->waitevent);
> + wait_event_timeout(msginfo->waitevent,
> + msginfo->wait_condition,
> + msecs_to_jiffies(1000));
> + BUG_ON(msginfo->wait_condition == 0);
The added BUG_ONs all over the code look scary. These shouldn't be
BUG_ONs at all. You should maybe warn and bail out, but not kill the
whole machine.
And looking at the code, more appropriate would be completion instead of
wait events.
And msecs_to_jiffies(1000) == HZ.
> @@ -689,7 +693,8 @@ static void vmbus_ongpadl_torndown(
> memcpy(&msginfo->response.gpadl_torndown,
> gpadl_torndown,
> sizeof(struct vmbus_channel_gpadl_torndown));
> - osd_waitevent_set(msginfo->waitevent);
> + msginfo->wait_condition = 1;
> + wake_up(&msginfo->waitevent);
> break;
> }
> }
> @@ -730,7 +735,8 @@ static void vmbus_onversion_response(
> memcpy(&msginfo->response.version_response,
> version_response,
> sizeof(struct vmbus_channel_version_response));
> - osd_waitevent_set(msginfo->waitevent);
> + msginfo->wait_condition = 1;
> + wake_up(&msginfo->waitevent);
> }
> }
> spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
regards,
--
js
> -----Original Message-----
> From: Jiri Slaby [mailto:[email protected]]
> Sent: Tuesday, February 15, 2011 4:21 AM
> To: KY Srinivasan
> Cc: [email protected]; [email protected];
> [email protected]; [email protected]
> Subject: Re: [PATCH 2/3]: Staging: hv: Use native wait primitives
>
> On 02/11/2011 06:59 PM, K. Y. Srinivasan wrote:
> > In preperation for getting rid of the osd layer; change
> > the code to use native wait interfaces. As part of this,
> > fixed the buggy implementation in the osd_wait_primitive
> > where the condition was cleared potentially after the
> > condition was signalled.
> ...
> > @@ -566,7 +567,11 @@ int vmbus_establish_gpadl(struct vmbus_channel
> *channel, void *kbuffer,
> >
> > }
> > }
> > - osd_waitevent_wait(msginfo->waitevent);
> > + wait_event_timeout(msginfo->waitevent,
> > + msginfo->wait_condition,
> > + msecs_to_jiffies(1000));
> > + BUG_ON(msginfo->wait_condition == 0);
>
> The added BUG_ONs all over the code look scary. These shouldn't be
> BUG_ONs at all. You should maybe warn and bail out, but not kill the
> whole machine.
This is Linux code running as a guest on a Windows host; and so the guest cannot
tolerate a failure of the host. In the cases where I have chosen to BUG_ON, there
is no reasonable recovery possible when the host is non-functional (as determined
by a non-responsive host).
>
> And looking at the code, more appropriate would be completion instead of
> wait events.
>
> And msecs_to_jiffies(1000) == HZ.
Agreed. In this first round of cleanup, I chose to keep the primitives as they were in osd.c. Greg, if it is ok with you, I will send you a patch that fixes these issues on top of the patches I have already sent.
Regards,
K. Y
>
> > @@ -689,7 +693,8 @@ static void vmbus_ongpadl_torndown(
> > memcpy(&msginfo->response.gpadl_torndown,
> > gpadl_torndown,
> > sizeof(struct
> vmbus_channel_gpadl_torndown));
> > - osd_waitevent_set(msginfo->waitevent);
> > + msginfo->wait_condition = 1;
> > + wake_up(&msginfo->waitevent);
> > break;
> > }
> > }
> > @@ -730,7 +735,8 @@ static void vmbus_onversion_response(
> > memcpy(&msginfo->response.version_response,
> > version_response,
> > sizeof(struct vmbus_channel_version_response));
> > - osd_waitevent_set(msginfo->waitevent);
> > + msginfo->wait_condition = 1;
> > + wake_up(&msginfo->waitevent);
> > }
> > }
> > spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
>
> regards,
> --
> js
On Tue, Feb 15, 2011 at 01:35:56PM +0000, KY Srinivasan wrote:
>
>
> > -----Original Message-----
> > From: Jiri Slaby [mailto:[email protected]]
> > Sent: Tuesday, February 15, 2011 4:21 AM
> > To: KY Srinivasan
> > Cc: [email protected]; [email protected];
> > [email protected]; [email protected]
> > Subject: Re: [PATCH 2/3]: Staging: hv: Use native wait primitives
> >
> > On 02/11/2011 06:59 PM, K. Y. Srinivasan wrote:
> > > In preperation for getting rid of the osd layer; change
> > > the code to use native wait interfaces. As part of this,
> > > fixed the buggy implementation in the osd_wait_primitive
> > > where the condition was cleared potentially after the
> > > condition was signalled.
> > ...
> > > @@ -566,7 +567,11 @@ int vmbus_establish_gpadl(struct vmbus_channel
> > *channel, void *kbuffer,
> > >
> > > }
> > > }
> > > - osd_waitevent_wait(msginfo->waitevent);
> > > + wait_event_timeout(msginfo->waitevent,
> > > + msginfo->wait_condition,
> > > + msecs_to_jiffies(1000));
> > > + BUG_ON(msginfo->wait_condition == 0);
> >
> > The added BUG_ONs all over the code look scary. These shouldn't be
> > BUG_ONs at all. You should maybe warn and bail out, but not kill the
> > whole machine.
>
> This is Linux code running as a guest on a Windows host; and so the guest cannot
> tolerate a failure of the host. In the cases where I have chosen to BUG_ON, there
> is no reasonable recovery possible when the host is non-functional (as determined
> by a non-responsive host).
If you have a non-responsive host, wouldn't that imply that this guest
code wouldn't run at all? :)
Having BUG_ON() in drivers is not a good idea either way. Please remove
these in future patches.
> > And looking at the code, more appropriate would be completion instead of
> > wait events.
> >
> > And msecs_to_jiffies(1000) == HZ.
>
> Agreed. In this first round of cleanup, I chose to keep the primitives
> as they were in osd.c. Greg, if it is ok with you, I will send you a
> patch that fixes these issues on top of the patches I have already
> sent.
Yes, that is fine.
thanks,
greg k-h
> -----Original Message-----
> From: Greg KH [mailto:[email protected]]
> Sent: Tuesday, February 15, 2011 9:03 AM
> To: KY Srinivasan
> Cc: Jiri Slaby; [email protected]; [email protected];
> [email protected]
> Subject: Re: [PATCH 2/3]: Staging: hv: Use native wait primitives
>
> On Tue, Feb 15, 2011 at 01:35:56PM +0000, KY Srinivasan wrote:
> >
> >
> > > -----Original Message-----
> > > From: Jiri Slaby [mailto:[email protected]]
> > > Sent: Tuesday, February 15, 2011 4:21 AM
> > > To: KY Srinivasan
> > > Cc: [email protected]; [email protected];
> > > [email protected]; [email protected]
> > > Subject: Re: [PATCH 2/3]: Staging: hv: Use native wait primitives
> > >
> > > On 02/11/2011 06:59 PM, K. Y. Srinivasan wrote:
> > > > In preperation for getting rid of the osd layer; change
> > > > the code to use native wait interfaces. As part of this,
> > > > fixed the buggy implementation in the osd_wait_primitive
> > > > where the condition was cleared potentially after the
> > > > condition was signalled.
> > > ...
> > > > @@ -566,7 +567,11 @@ int vmbus_establish_gpadl(struct vmbus_channel
> > > *channel, void *kbuffer,
> > > >
> > > > }
> > > > }
> > > > - osd_waitevent_wait(msginfo->waitevent);
> > > > + wait_event_timeout(msginfo->waitevent,
> > > > + msginfo->wait_condition,
> > > > + msecs_to_jiffies(1000));
> > > > + BUG_ON(msginfo->wait_condition == 0);
> > >
> > > The added BUG_ONs all over the code look scary. These shouldn't be
> > > BUG_ONs at all. You should maybe warn and bail out, but not kill the
> > > whole machine.
> >
> > This is Linux code running as a guest on a Windows host; and so the guest
> cannot
> > tolerate a failure of the host. In the cases where I have chosen to BUG_ON,
> there
> > is no reasonable recovery possible when the host is non-functional (as
> determined
> > by a non-responsive host).
>
> If you have a non-responsive host, wouldn't that imply that this guest
> code wouldn't run at all? :)
The fact that on a particular transaction the host has not responded within an expected
time interval does not necessarily mean that the guest code would not be running. There may be
issues on the host side that may be either transient or permanent that may cause problems like
this. Keep in mind, HyperV is a type 1 hypervisor that would schedule all VMs including the host
and so, guest would get scheduled.
>
> Having BUG_ON() in drivers is not a good idea either way. Please remove
> these in future patches.
In situations where there is not a reasonable rollback strategy (for
instance in one of the cases, we are granting access to the guest
physical pages to the host) we really have only 2 options:
1) Wait until the host responds. This wait could potentially be unbounded
and in fact this was the way the code was to begin with. One of the reviewers
had suggested that unbounded wait was to be corrected.
2) Wait for a specific period and if the host does not respond
within a reasonable period, kill the guest since there is no recovery
possible.
I chose option 2, as part of addressing some of the prior review
comments. If the consensus now is to go back to option 1, I am fine with that;
I will send you a patch to rectify this.
Regards,
K. Y
>
> > > And looking at the code, more appropriate would be completion instead of
> > > wait events.
> > >
> > > And msecs_to_jiffies(1000) == HZ.
> >
> > Agreed. In this first round of cleanup, I chose to keep the primitives
> > as they were in osd.c. Greg, if it is ok with you, I will send you a
> > patch that fixes these issues on top of the patches I have already
> > sent.
>
> Yes, that is fine.
>
> thanks,
>
> greg k-h
On Tue, Feb 15, 2011 at 04:22:20PM +0000, KY Srinivasan wrote:
>
>
> > -----Original Message-----
> > From: Greg KH [mailto:[email protected]]
> > Sent: Tuesday, February 15, 2011 9:03 AM
> > To: KY Srinivasan
> > Cc: Jiri Slaby; [email protected]; [email protected];
> > [email protected]
> > Subject: Re: [PATCH 2/3]: Staging: hv: Use native wait primitives
> >
> > On Tue, Feb 15, 2011 at 01:35:56PM +0000, KY Srinivasan wrote:
> > >
> > >
> > > > -----Original Message-----
> > > > From: Jiri Slaby [mailto:[email protected]]
> > > > Sent: Tuesday, February 15, 2011 4:21 AM
> > > > To: KY Srinivasan
> > > > Cc: [email protected]; [email protected];
> > > > [email protected]; [email protected]
> > > > Subject: Re: [PATCH 2/3]: Staging: hv: Use native wait primitives
> > > >
> > > > On 02/11/2011 06:59 PM, K. Y. Srinivasan wrote:
> > > > > In preperation for getting rid of the osd layer; change
> > > > > the code to use native wait interfaces. As part of this,
> > > > > fixed the buggy implementation in the osd_wait_primitive
> > > > > where the condition was cleared potentially after the
> > > > > condition was signalled.
> > > > ...
> > > > > @@ -566,7 +567,11 @@ int vmbus_establish_gpadl(struct vmbus_channel
> > > > *channel, void *kbuffer,
> > > > >
> > > > > }
> > > > > }
> > > > > - osd_waitevent_wait(msginfo->waitevent);
> > > > > + wait_event_timeout(msginfo->waitevent,
> > > > > + msginfo->wait_condition,
> > > > > + msecs_to_jiffies(1000));
> > > > > + BUG_ON(msginfo->wait_condition == 0);
> > > >
> > > > The added BUG_ONs all over the code look scary. These shouldn't be
> > > > BUG_ONs at all. You should maybe warn and bail out, but not kill the
> > > > whole machine.
> > >
> > > This is Linux code running as a guest on a Windows host; and so the guest
> > cannot
> > > tolerate a failure of the host. In the cases where I have chosen to BUG_ON,
> > there
> > > is no reasonable recovery possible when the host is non-functional (as
> > determined
> > > by a non-responsive host).
> >
> > If you have a non-responsive host, wouldn't that imply that this guest
> > code wouldn't run at all? :)
>
> The fact that on a particular transaction the host has not responded within an expected
> time interval does not necessarily mean that the guest code would not be running. There may be
> issues on the host side that may be either transient or permanent that may cause problems like
> this. Keep in mind, HyperV is a type 1 hypervisor that would schedule all VMs including the host
> and so, guest would get scheduled.
>
> >
> > Having BUG_ON() in drivers is not a good idea either way. Please remove
> > these in future patches.
>
> In situations where there is not a reasonable rollback strategy (for
> instance in one of the cases, we are granting access to the guest
> physical pages to the host) we really have only 2 options:
>
> 1) Wait until the host responds. This wait could potentially be unbounded
> and in fact this was the way the code was to begin with. One of the reviewers
> had suggested that unbounded wait was to be corrected.
> 2) Wait for a specific period and if the host does not respond
> within a reasonable period, kill the guest since there is no recovery
> possible.
Killing the guest is a very serious thing, causing all sorts of possible
problems with it, right?
> I chose option 2, as part of addressing some of the prior review
> comments. If the consensus now is to go back to option 1, I am fine with that;
Unbounded waits aren't ok either, you need some sort of timeout.
But, as this is a bit preferable to dieing, I suggest doing this, and
comment the heck out of it to explain all of this for anyone who reads
it.
thanks,
greg k-h
> -----Original Message-----
> From: Greg KH [mailto:[email protected]]
> Sent: Tuesday, February 15, 2011 11:30 AM
> To: KY Srinivasan
> Cc: Jiri Slaby; [email protected]; [email protected];
> [email protected]
> Subject: Re: [PATCH 2/3]: Staging: hv: Use native wait primitives
>
> On Tue, Feb 15, 2011 at 04:22:20PM +0000, KY Srinivasan wrote:
> >
> >
> > > -----Original Message-----
> > > From: Greg KH [mailto:[email protected]]
> > > Sent: Tuesday, February 15, 2011 9:03 AM
> > > To: KY Srinivasan
> > > Cc: Jiri Slaby; [email protected]; [email protected];
> > > [email protected]
> > > Subject: Re: [PATCH 2/3]: Staging: hv: Use native wait primitives
> > >
> > > On Tue, Feb 15, 2011 at 01:35:56PM +0000, KY Srinivasan wrote:
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Jiri Slaby [mailto:[email protected]]
> > > > > Sent: Tuesday, February 15, 2011 4:21 AM
> > > > > To: KY Srinivasan
> > > > > Cc: [email protected]; [email protected];
> > > > > [email protected]; [email protected]
> > > > > Subject: Re: [PATCH 2/3]: Staging: hv: Use native wait primitives
> > > > >
> > > > > On 02/11/2011 06:59 PM, K. Y. Srinivasan wrote:
> > > > > > In preperation for getting rid of the osd layer; change
> > > > > > the code to use native wait interfaces. As part of this,
> > > > > > fixed the buggy implementation in the osd_wait_primitive
> > > > > > where the condition was cleared potentially after the
> > > > > > condition was signalled.
> > > > > ...
> > > > > > @@ -566,7 +567,11 @@ int vmbus_establish_gpadl(struct
> vmbus_channel
> > > > > *channel, void *kbuffer,
> > > > > >
> > > > > > }
> > > > > > }
> > > > > > - osd_waitevent_wait(msginfo->waitevent);
> > > > > > + wait_event_timeout(msginfo->waitevent,
> > > > > > + msginfo->wait_condition,
> > > > > > + msecs_to_jiffies(1000));
> > > > > > + BUG_ON(msginfo->wait_condition == 0);
> > > > >
> > > > > The added BUG_ONs all over the code look scary. These shouldn't be
> > > > > BUG_ONs at all. You should maybe warn and bail out, but not kill the
> > > > > whole machine.
> > > >
> > > > This is Linux code running as a guest on a Windows host; and so the guest
> > > cannot
> > > > tolerate a failure of the host. In the cases where I have chosen to BUG_ON,
> > > there
> > > > is no reasonable recovery possible when the host is non-functional (as
> > > determined
> > > > by a non-responsive host).
> > >
> > > If you have a non-responsive host, wouldn't that imply that this guest
> > > code wouldn't run at all? :)
> >
> > The fact that on a particular transaction the host has not responded within an
> expected
> > time interval does not necessarily mean that the guest code would not be
> running. There may be
> > issues on the host side that may be either transient or permanent that may
> cause problems like
> > this. Keep in mind, HyperV is a type 1 hypervisor that would schedule all VMs
> including the host
> > and so, guest would get scheduled.
> >
> > >
> > > Having BUG_ON() in drivers is not a good idea either way. Please remove
> > > these in future patches.
> >
> > In situations where there is not a reasonable rollback strategy (for
> > instance in one of the cases, we are granting access to the guest
> > physical pages to the host) we really have only 2 options:
> >
> > 1) Wait until the host responds. This wait could potentially be unbounded
> > and in fact this was the way the code was to begin with. One of the reviewers
> > had suggested that unbounded wait was to be corrected.
> > 2) Wait for a specific period and if the host does not respond
> > within a reasonable period, kill the guest since there is no recovery
> > possible.
>
> Killing the guest is a very serious thing, causing all sorts of possible
> problems with it, right?
If there was a reasonable rollback strategy, I would not
be killing the guest.
>
> > I chose option 2, as part of addressing some of the prior review
> > comments. If the consensus now is to go back to option 1, I am fine with that;
>
> Unbounded waits aren't ok either, you need some sort of timeout.
>
> But, as this is a bit preferable to dieing, I suggest doing this, and
> comment the heck out of it to explain all of this for anyone who reads
> it.
If I understand you correctly, you would be prefer to have unbounded waiting with comments
justifying why we cannot have timeouts. I will roll out a patch once the tree stabilizes.
Regards,
K. Y
>
> thanks,
>
> greg k-h
On Tue, Feb 15, 2011 at 05:52:46PM +0000, KY Srinivasan wrote:
>
> If I understand you correctly, you would be prefer to have unbounded waiting with comments
> justifying why we cannot have timeouts. I will roll out a patch once the tree stabilizes.
>
Just make sure to add a comment explaining why the wait is unbounded.
--
Dmitry