github/linuxkit
1
0
Fork 0
mirror of https://github.com/linuxkit/linuxkit.git synced 2025-07-27 20:48:11 +00:00
Code Issues Projects Releases Wiki Activity

kernel: Add VMBus stability improvements

Browse Source 
The four new patches prevent swamping VMBus with too many
notifications which, eventually, let Hyper-V assume there is
a DoS attempt and would thus throttle the VM for up to 5s.

The first three new patches were cherry-picked from upstream.
The final patch was received by email. Will update once it
has been submitted/accepted upstream.

Generated from: https://github.com/rneugeba/linux-stable/tree/v4.9.5-moby

Signed-off-by: Rolf Neugebauer <rolf.neugebauer@docker.com>
This commit is contained in:
Rolf Neugebauer 2017-01-20 16:07:52 +00:00
parent 7f4d396ea6
commit e7575f88da
11 changed files with 715 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
alpine/kernel/patches-4.9/0001-VSOCK-Only-allow-host-network-namespace-to-use-AF_VS.patch
View File

@ -1,7 +1,8 @@
From 1fc22bc5c2e3544786420355405038e4a12ffd72 Mon Sep 17 00:00:00 2001
From: Ian Campbell <ian.campbell@docker.com>
Date: Mon, 4 Apr 2016 14:50:10 +0100
Subject: [PATCH 1/7] VSOCK: Only allow host network namespace to use AF_VSOCK.
Subject: [PATCH 01/11] VSOCK: Only allow host network namespace to use
AF_VSOCK.
The VSOCK addressing schema does not really lend itself to simply creating an
alternative end point address within a namespace.

2
alpine/kernel/patches-4.9/0002-hv_sock-introduce-Hyper-V-Sockets.patch
View File

@ -1,7 +1,7 @@
From 0d9748d2d26216421225b41643c6167fda91c26f Mon Sep 17 00:00:00 2001
From: Dexuan Cui <decui@microsoft.com>
Date: Sat, 23 Jul 2016 01:35:51 +0000
Subject: [PATCH 2/7] hv_sock: introduce Hyper-V Sockets
Subject: [PATCH 02/11] hv_sock: introduce Hyper-V Sockets
Hyper-V Sockets (hv_sock) supplies a byte-stream based communication
mechanism between the host and the guest. It's somewhat like TCP over

2
alpine/kernel/patches-4.9/0003-vmbus-Don-t-spam-the-logs-with-unknown-GUIDs.patch
View File

@ -1,7 +1,7 @@
From c9168a8ff6c84924df9efec05257f74ed93981b2 Mon Sep 17 00:00:00 2001
From: Rolf Neugebauer <rolf.neugebauer@gmail.com>
Date: Mon, 23 May 2016 18:55:45 +0100
Subject: [PATCH 3/7] vmbus: Don't spam the logs with unknown GUIDs
Subject: [PATCH 03/11] vmbus: Don't spam the logs with unknown GUIDs
With Hyper-V sockets device types are introduced on the fly. The pr_info()
then prints a message on every connection, which is way too verbose. Since

4
alpine/kernel/patches-4.9/0004-Drivers-hv-utils-Fix-the-mapping-between-host-versio.patch
View File

@ -1,8 +1,8 @@
From 24a17904f276ced95d717fc427c4f6ff256ac0a8 Mon Sep 17 00:00:00 2001
From: Alex Ng <alexng@messages.microsoft.com>
Date: Sun, 6 Nov 2016 13:14:07 -0800
Subject: [PATCH 4/7] Drivers: hv: utils: Fix the mapping between host version
and protocol to use
Subject: [PATCH 04/11] Drivers: hv: utils: Fix the mapping between host
version and protocol to use
We should intentionally declare the protocols to use for every known host
and default to using the latest protocol if the host is unknown or new.

4
alpine/kernel/patches-4.9/0005-Drivers-hv-utils-Force-TimeSync-version-3.0-on-Windo.patch
View File

@ -1,8 +1,8 @@
From 5d339781f71a37083d4c4d304ba45b89aede184f Mon Sep 17 00:00:00 2001
From: Rolf Neugebauer <rolf.neugebauer@docker.com>
Date: Wed, 11 Jan 2017 22:40:38 +0000
Subject: [PATCH 5/7] Drivers: hv: utils: Force TimeSync version 3.0 on Windows
10
Subject: [PATCH 05/11] Drivers: hv: utils: Force TimeSync version 3.0 on
Windows 10
Some older Windows 10 builds, including 10586 do not seem to
support TimeSync protocol 4.0 causing loss of time synchronisation

2
alpine/kernel/patches-4.9/0006-fuse-fix-time_to_jiffies-nsec-sanity-check.patch
View File

@ -1,7 +1,7 @@
From 4c63fbce6abc0eb36a21d760959aae954427dcf3 Mon Sep 17 00:00:00 2001
From: David Sheets <david.sheets@docker.com>
Date: Fri, 13 Jan 2017 15:58:30 +0000
Subject: [PATCH 6/7] fuse: fix time_to_jiffies nsec sanity check
Subject: [PATCH 06/11] fuse: fix time_to_jiffies nsec sanity check
Commit bcb6f6d2b9c2 ("fuse: use timespec64") introduced clamped nsec values
in time_to_jiffies but used the max of nsec and NSEC_PER_SEC - 1 instead of

2
alpine/kernel/patches-4.9/0007-virtio-don-t-set-VIRTIO_NET_HDR_F_DATA_VALID-on-xmit.patch
View File

@ -1,7 +1,7 @@
From 00ef9fe4743da2cdd5b55d97335e678304329e24 Mon Sep 17 00:00:00 2001
From: Rolf Neugebauer <rolf.neugebauer@docker.com>
Date: Tue, 17 Jan 2017 18:13:51 +0000
Subject: [PATCH 7/7] virtio: don't set VIRTIO_NET_HDR_F_DATA_VALID on xmit
Subject: [PATCH 07/11] virtio: don't set VIRTIO_NET_HDR_F_DATA_VALID on xmit
This patch part reverts fd2a0437dc33 and e858fae2b0b8 which introduced a
subtle change in how the virtio_net flags are derived from the SKBs

103
alpine/kernel/patches-4.9/0008-Drivers-hv-vmbus-Base-host-signaling-strictly-on-the.patch Normal file
View File

@ -0,0 +1,103 @@
From 72c60164c3508d229686d64e624d1d2f47d01676 Mon Sep 17 00:00:00 2001
From: "K. Y. Srinivasan" <kys@microsoft.com>
Date: Sun, 6 Nov 2016 13:14:16 -0800
Subject: [PATCH 08/11] Drivers: hv: vmbus: Base host signaling strictly on the
ring state
One of the factors that can result in the host concluding that a given
guest in mounting a DOS attack is if the guest generates interrupts
to the host when the host is not expecting it. If these "spurious"
interrupts reach a certain rate, the host can throttle the guest to
minimize the impact. The host computation of the "expected number
of interrupts" is strictly based on the ring transitions. Until
the host logic is fixed, base the guest logic to interrupt solely
on the ring state.
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 74198eb4a42c4a3c4fbef08fa01a291a282f7c2e)
---
drivers/hv/channel.c | 23 ++++++++++++++++++++---
drivers/hv/channel_mgmt.c | 2 --
drivers/hv/ring_buffer.c | 7 -------
3 files changed, 20 insertions(+), 12 deletions(-)
diff --git a/drivers/hv/channel.c b/drivers/hv/channel.c
index 16f91c8490fe..5e482d7f60cb 100644
--- a/drivers/hv/channel.c
+++ b/drivers/hv/channel.c
@@ -676,10 +676,18 @@ int vmbus_sendpacket_ctl(struct vmbus_channel *channel, void *buffer,
* NOTE: in this case, the hvsock channel is an exception, because
* it looks the host side's hvsock implementation has a throttling
* mechanism which can hurt the performance otherwise.
+ *
+ * KYS: Oct. 30, 2016:
+ * It looks like Windows hosts have logic to deal with DOS attacks that
+ * can be triggered if it receives interrupts when it is not expecting
+ * the interrupt. The host expects interrupts only when the ring
+ * transitions from empty to non-empty (or full to non full on the guest
+ * to host ring).
+ * So, base the signaling decision solely on the ring state until the
+ * host logic is fixed.
*/
- if (((ret == 0) && kick_q && signal) ||
- (ret && !is_hvsock_channel(channel)))
+ if (((ret == 0) && signal))
vmbus_setevent(channel);
return ret;
@@ -786,9 +794,18 @@ int vmbus_sendpacket_pagebuffer_ctl(struct vmbus_channel *channel,
* If we cannot write to the ring-buffer; signal the host
* even if we may not have written anything. This is a rare
* enough condition that it should not matter.
+ *
+ * KYS: Oct. 30, 2016:
+ * It looks like Windows hosts have logic to deal with DOS attacks that
+ * can be triggered if it receives interrupts when it is not expecting
+ * the interrupt. The host expects interrupts only when the ring
+ * transitions from empty to non-empty (or full to non full on the guest
+ * to host ring).
+ * So, base the signaling decision solely on the ring state until the
+ * host logic is fixed.
*/
- if (((ret == 0) && kick_q && signal) || (ret))
+ if (((ret == 0) && signal))
vmbus_setevent(channel);
return ret;
diff --git a/drivers/hv/channel_mgmt.c b/drivers/hv/channel_mgmt.c
index db5bccf4fa7e..8f3d9f787288 100644
--- a/drivers/hv/channel_mgmt.c
+++ b/drivers/hv/channel_mgmt.c
@@ -448,8 +448,6 @@ static void vmbus_process_offer(struct vmbus_channel *newchannel)
}
dev_type = hv_get_dev_type(newchannel);
- if (dev_type == HV_NIC)
- set_channel_signal_state(newchannel, HV_SIGNAL_POLICY_EXPLICIT);
init_vp_index(newchannel, dev_type);
diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index 08043da1a61c..5d11d93eedf4 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c
@@ -75,13 +75,6 @@ static bool hv_need_to_signal(u32 old_write, struct hv_ring_buffer_info *rbi,
if (READ_ONCE(rbi->ring_buffer->interrupt_mask))
return false;
- /*
- * When the client wants to control signaling,
- * we only honour the host interrupt mask.
- */
- if (policy == HV_SIGNAL_POLICY_EXPLICIT)
- return true;
-
/* check interrupt_mask before read_index */
virt_rmb();
/*
--
2.11.0

321
alpine/kernel/patches-4.9/0009-Drivers-hv-vmbus-On-write-cleanup-the-logic-to-inter.patch Normal file
View File

@ -0,0 +1,321 @@
From f0472d013423117cae77e00089485f7d976138f2 Mon Sep 17 00:00:00 2001
From: "K. Y. Srinivasan" <kys@microsoft.com>
Date: Sun, 6 Nov 2016 13:14:17 -0800
Subject: [PATCH 09/11] Drivers: hv: vmbus: On write cleanup the logic to
interrupt the host
Signal the host when we determine the host is to be signaled.
The currrent code determines the need to signal in the ringbuffer
code and actually issues the signal elsewhere. This can result
in the host viewing this interrupt as spurious since the host may also
poll the channel. Make the necessary adjustments.
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 1f6ee4e7d83586c8b10bd4f2f4346353d04ce884)
---
drivers/hv/channel.c | 99 +++++------------------------------------------
drivers/hv/hyperv_vmbus.h | 6 +--
drivers/hv/ring_buffer.c | 30 +++++++++-----
include/linux/hyperv.h | 1 +
4 files changed, 35 insertions(+), 101 deletions(-)
diff --git a/drivers/hv/channel.c b/drivers/hv/channel.c
index 5e482d7f60cb..8a8148f7b842 100644
--- a/drivers/hv/channel.c
+++ b/drivers/hv/channel.c
@@ -39,7 +39,7 @@
* vmbus_setevent- Trigger an event notification on the specified
* channel.
*/
-static void vmbus_setevent(struct vmbus_channel *channel)
+void vmbus_setevent(struct vmbus_channel *channel)
{
struct hv_monitor_page *monitorpage;
@@ -65,6 +65,7 @@ static void vmbus_setevent(struct vmbus_channel *channel)
vmbus_set_event(channel);
}
}
+EXPORT_SYMBOL_GPL(vmbus_setevent);
/*
* vmbus_open - Open the specified channel.
@@ -635,8 +636,6 @@ int vmbus_sendpacket_ctl(struct vmbus_channel *channel, void *buffer,
u32 packetlen_aligned = ALIGN(packetlen, sizeof(u64));
struct kvec bufferlist[3];
u64 aligned_data = 0;
- int ret;
- bool signal = false;
bool lock = channel->acquire_ring_lock;
int num_vecs = ((bufferlen != 0) ? 3 : 1);
@@ -656,41 +655,9 @@ int vmbus_sendpacket_ctl(struct vmbus_channel *channel, void *buffer,
bufferlist[2].iov_base = &aligned_data;
bufferlist[2].iov_len = (packetlen_aligned - packetlen);
- ret = hv_ringbuffer_write(&channel->outbound, bufferlist, num_vecs,
- &signal, lock, channel->signal_policy);
-
- /*
- * Signalling the host is conditional on many factors:
- * 1. The ring state changed from being empty to non-empty.
- * This is tracked by the variable "signal".
- * 2. The variable kick_q tracks if more data will be placed
- * on the ring. We will not signal if more data is
- * to be placed.
- *
- * Based on the channel signal state, we will decide
- * which signaling policy will be applied.
- *
- * If we cannot write to the ring-buffer; signal the host
- * even if we may not have written anything. This is a rare
- * enough condition that it should not matter.
- * NOTE: in this case, the hvsock channel is an exception, because
- * it looks the host side's hvsock implementation has a throttling
- * mechanism which can hurt the performance otherwise.
- *
- * KYS: Oct. 30, 2016:
- * It looks like Windows hosts have logic to deal with DOS attacks that
- * can be triggered if it receives interrupts when it is not expecting
- * the interrupt. The host expects interrupts only when the ring
- * transitions from empty to non-empty (or full to non full on the guest
- * to host ring).
- * So, base the signaling decision solely on the ring state until the
- * host logic is fixed.
- */
-
- if (((ret == 0) && signal))
- vmbus_setevent(channel);
+ return hv_ringbuffer_write(channel, bufferlist, num_vecs,
+ lock, kick_q);
- return ret;
}
EXPORT_SYMBOL(vmbus_sendpacket_ctl);
@@ -731,7 +698,6 @@ int vmbus_sendpacket_pagebuffer_ctl(struct vmbus_channel *channel,
u32 flags,
bool kick_q)
{
- int ret;
int i;
struct vmbus_channel_packet_page_buffer desc;
u32 descsize;
@@ -739,7 +705,6 @@ int vmbus_sendpacket_pagebuffer_ctl(struct vmbus_channel *channel,
u32 packetlen_aligned;
struct kvec bufferlist[3];
u64 aligned_data = 0;
- bool signal = false;
bool lock = channel->acquire_ring_lock;
if (pagecount > MAX_PAGE_BUFFER_COUNT)
@@ -777,38 +742,8 @@ int vmbus_sendpacket_pagebuffer_ctl(struct vmbus_channel *channel,
bufferlist[2].iov_base = &aligned_data;
bufferlist[2].iov_len = (packetlen_aligned - packetlen);
- ret = hv_ringbuffer_write(&channel->outbound, bufferlist, 3,
- &signal, lock, channel->signal_policy);
-
- /*
- * Signalling the host is conditional on many factors:
- * 1. The ring state changed from being empty to non-empty.
- * This is tracked by the variable "signal".
- * 2. The variable kick_q tracks if more data will be placed
- * on the ring. We will not signal if more data is
- * to be placed.
- *
- * Based on the channel signal state, we will decide
- * which signaling policy will be applied.
- *
- * If we cannot write to the ring-buffer; signal the host
- * even if we may not have written anything. This is a rare
- * enough condition that it should not matter.
- *
- * KYS: Oct. 30, 2016:
- * It looks like Windows hosts have logic to deal with DOS attacks that
- * can be triggered if it receives interrupts when it is not expecting
- * the interrupt. The host expects interrupts only when the ring
- * transitions from empty to non-empty (or full to non full on the guest
- * to host ring).
- * So, base the signaling decision solely on the ring state until the
- * host logic is fixed.
- */
-
- if (((ret == 0) && signal))
- vmbus_setevent(channel);
-
- return ret;
+ return hv_ringbuffer_write(channel, bufferlist, 3,
+ lock, kick_q);
}
EXPORT_SYMBOL_GPL(vmbus_sendpacket_pagebuffer_ctl);
@@ -839,12 +774,10 @@ int vmbus_sendpacket_mpb_desc(struct vmbus_channel *channel,
u32 desc_size,
void *buffer, u32 bufferlen, u64 requestid)
{
- int ret;
u32 packetlen;
u32 packetlen_aligned;
struct kvec bufferlist[3];
u64 aligned_data = 0;
- bool signal = false;
bool lock = channel->acquire_ring_lock;
packetlen = desc_size + bufferlen;
@@ -865,13 +798,8 @@ int vmbus_sendpacket_mpb_desc(struct vmbus_channel *channel,
bufferlist[2].iov_base = &aligned_data;
bufferlist[2].iov_len = (packetlen_aligned - packetlen);
- ret = hv_ringbuffer_write(&channel->outbound, bufferlist, 3,
- &signal, lock, channel->signal_policy);
-
- if (ret == 0 && signal)
- vmbus_setevent(channel);
-
- return ret;
+ return hv_ringbuffer_write(channel, bufferlist, 3,
+ lock, true);
}
EXPORT_SYMBOL_GPL(vmbus_sendpacket_mpb_desc);
@@ -883,14 +811,12 @@ int vmbus_sendpacket_multipagebuffer(struct vmbus_channel *channel,
struct hv_multipage_buffer *multi_pagebuffer,
void *buffer, u32 bufferlen, u64 requestid)
{
- int ret;
struct vmbus_channel_packet_multipage_buffer desc;
u32 descsize;
u32 packetlen;
u32 packetlen_aligned;
struct kvec bufferlist[3];
u64 aligned_data = 0;
- bool signal = false;
bool lock = channel->acquire_ring_lock;
u32 pfncount = NUM_PAGES_SPANNED(multi_pagebuffer->offset,
multi_pagebuffer->len);
@@ -930,13 +856,8 @@ int vmbus_sendpacket_multipagebuffer(struct vmbus_channel *channel,
bufferlist[2].iov_base = &aligned_data;
bufferlist[2].iov_len = (packetlen_aligned - packetlen);
- ret = hv_ringbuffer_write(&channel->outbound, bufferlist, 3,
- &signal, lock, channel->signal_policy);
-
- if (ret == 0 && signal)
- vmbus_setevent(channel);
-
- return ret;
+ return hv_ringbuffer_write(channel, bufferlist, 3,
+ lock, true);
}
EXPORT_SYMBOL_GPL(vmbus_sendpacket_multipagebuffer);
diff --git a/drivers/hv/hyperv_vmbus.h b/drivers/hv/hyperv_vmbus.h
index a5b4442433c8..fa782e13c8da 100644
--- a/drivers/hv/hyperv_vmbus.h
+++ b/drivers/hv/hyperv_vmbus.h
@@ -527,10 +527,10 @@ int hv_ringbuffer_init(struct hv_ring_buffer_info *ring_info,
void hv_ringbuffer_cleanup(struct hv_ring_buffer_info *ring_info);
-int hv_ringbuffer_write(struct hv_ring_buffer_info *ring_info,
+int hv_ringbuffer_write(struct vmbus_channel *channel,
struct kvec *kv_list,
- u32 kv_count, bool *signal, bool lock,
- enum hv_signal_policy policy);
+ u32 kv_count, bool lock,
+ bool kick_q);
int hv_ringbuffer_read(struct hv_ring_buffer_info *inring_info,
void *buffer, u32 buflen, u32 *buffer_actual_len,
diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index 5d11d93eedf4..4af71306d0ff 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c
@@ -66,14 +66,25 @@ u32 hv_end_read(struct hv_ring_buffer_info *rbi)
* once the ring buffer is empty, it will clear the
* interrupt_mask and re-check to see if new data has
* arrived.
+ *
+ * KYS: Oct. 30, 2016:
+ * It looks like Windows hosts have logic to deal with DOS attacks that
+ * can be triggered if it receives interrupts when it is not expecting
+ * the interrupt. The host expects interrupts only when the ring
+ * transitions from empty to non-empty (or full to non full on the guest
+ * to host ring).
+ * So, base the signaling decision solely on the ring state until the
+ * host logic is fixed.
*/
-static bool hv_need_to_signal(u32 old_write, struct hv_ring_buffer_info *rbi,
- enum hv_signal_policy policy)
+static void hv_signal_on_write(u32 old_write, struct vmbus_channel *channel,
+ bool kick_q)
{
+ struct hv_ring_buffer_info *rbi = &channel->outbound;
+
virt_mb();
if (READ_ONCE(rbi->ring_buffer->interrupt_mask))
- return false;
+ return;
/* check interrupt_mask before read_index */
virt_rmb();
@@ -82,9 +93,9 @@ static bool hv_need_to_signal(u32 old_write, struct hv_ring_buffer_info *rbi,
* ring transitions from being empty to non-empty.
*/
if (old_write == READ_ONCE(rbi->ring_buffer->read_index))
- return true;
+ vmbus_setevent(channel);
- return false;
+ return;
}
/* Get the next write location for the specified ring buffer. */
@@ -273,9 +284,9 @@ void hv_ringbuffer_cleanup(struct hv_ring_buffer_info *ring_info)
}
/* Write to the ring buffer. */
-int hv_ringbuffer_write(struct hv_ring_buffer_info *outring_info,
- struct kvec *kv_list, u32 kv_count, bool *signal, bool lock,
- enum hv_signal_policy policy)
+int hv_ringbuffer_write(struct vmbus_channel *channel,
+ struct kvec *kv_list, u32 kv_count, bool lock,
+ bool kick_q)
{
int i = 0;
u32 bytes_avail_towrite;
@@ -285,6 +296,7 @@ int hv_ringbuffer_write(struct hv_ring_buffer_info *outring_info,
u32 old_write;
u64 prev_indices = 0;
unsigned long flags = 0;
+ struct hv_ring_buffer_info *outring_info = &channel->outbound;
for (i = 0; i < kv_count; i++)
totalbytes_towrite += kv_list[i].iov_len;
@@ -337,7 +349,7 @@ int hv_ringbuffer_write(struct hv_ring_buffer_info *outring_info,
if (lock)
spin_unlock_irqrestore(&outring_info->ring_lock, flags);
- *signal = hv_need_to_signal(old_write, outring_info, policy);
+ hv_signal_on_write(old_write, channel, kick_q);
return 0;
}
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index 468e15e29f5f..9a2d657a7fe7 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -1447,6 +1447,7 @@ void hv_event_tasklet_enable(struct vmbus_channel *channel);
void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid);
+void vmbus_setevent(struct vmbus_channel *channel);
/*
* Negotiated version with the Host.
*/
--
2.11.0

141
alpine/kernel/patches-4.9/0010-Drivers-hv-vmbus-On-the-read-path-cleanup-the-logic-.patch Normal file
View File

@ -0,0 +1,141 @@
From 9e0ab3ccac70147f6e0e04cf0872d6f477ec6b68 Mon Sep 17 00:00:00 2001
From: "K. Y. Srinivasan" <kys@microsoft.com>
Date: Sun, 6 Nov 2016 13:14:18 -0800
Subject: [PATCH 10/11] Drivers: hv: vmbus: On the read path cleanup the logic
to interrupt the host
Signal the host when we determine the host is to be signaled -
on th read path. The currrent code determines the need to signal in the
ringbuffer code and actually issues the signal elsewhere. This can result
in the host viewing this interrupt as spurious since the host may also
poll the channel. Make the necessary adjustments.
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
drivers/hv/channel.c | 11 ++---------
drivers/hv/hyperv_vmbus.h | 4 ++--
drivers/hv/ring_buffer.c | 7 ++++---
include/linux/hyperv.h | 12 ++++++------
4 files changed, 14 insertions(+), 20 deletions(-)
diff --git a/drivers/hv/channel.c b/drivers/hv/channel.c
index 8a8148f7b842..5fb4c6d9209b 100644
--- a/drivers/hv/channel.c
+++ b/drivers/hv/channel.c
@@ -879,16 +879,9 @@ __vmbus_recvpacket(struct vmbus_channel *channel, void *buffer,
u32 bufferlen, u32 *buffer_actual_len, u64 *requestid,
bool raw)
{
- int ret;
- bool signal = false;
-
- ret = hv_ringbuffer_read(&channel->inbound, buffer, bufferlen,
- buffer_actual_len, requestid, &signal, raw);
+ return hv_ringbuffer_read(channel, buffer, bufferlen,
+ buffer_actual_len, requestid, raw);
- if (signal)
- vmbus_setevent(channel);
-
- return ret;
}
int vmbus_recvpacket(struct vmbus_channel *channel, void *buffer,
diff --git a/drivers/hv/hyperv_vmbus.h b/drivers/hv/hyperv_vmbus.h
index fa782e13c8da..2b13f2a0a71e 100644
--- a/drivers/hv/hyperv_vmbus.h
+++ b/drivers/hv/hyperv_vmbus.h
@@ -532,9 +532,9 @@ int hv_ringbuffer_write(struct vmbus_channel *channel,
u32 kv_count, bool lock,
bool kick_q);
-int hv_ringbuffer_read(struct hv_ring_buffer_info *inring_info,
+int hv_ringbuffer_read(struct vmbus_channel *channel,
void *buffer, u32 buflen, u32 *buffer_actual_len,
- u64 *requestid, bool *signal, bool raw);
+ u64 *requestid, bool raw);
void hv_ringbuffer_get_debuginfo(struct hv_ring_buffer_info *ring_info,
struct hv_ring_buffer_debug_info *debug_info);
diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index 4af71306d0ff..cd49cb17eb7f 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c
@@ -353,9 +353,9 @@ int hv_ringbuffer_write(struct vmbus_channel *channel,
return 0;
}
-int hv_ringbuffer_read(struct hv_ring_buffer_info *inring_info,
+int hv_ringbuffer_read(struct vmbus_channel *channel,
void *buffer, u32 buflen, u32 *buffer_actual_len,
- u64 *requestid, bool *signal, bool raw)
+ u64 *requestid, bool raw)
{
u32 bytes_avail_toread;
u32 next_read_location = 0;
@@ -364,6 +364,7 @@ int hv_ringbuffer_read(struct hv_ring_buffer_info *inring_info,
u32 offset;
u32 packetlen;
int ret = 0;
+ struct hv_ring_buffer_info *inring_info = &channel->inbound;
if (buflen <= 0)
return -EINVAL;
@@ -421,7 +422,7 @@ int hv_ringbuffer_read(struct hv_ring_buffer_info *inring_info,
/* Update the read index */
hv_set_next_read_location(inring_info, next_read_location);
- *signal = hv_need_to_signal_on_read(inring_info);
+ hv_signal_on_read(channel);
return ret;
}
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index 9a2d657a7fe7..5343b7406acc 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -1480,10 +1480,11 @@ hv_get_ring_buffer(struct hv_ring_buffer_info *ring_info)
* there is room for the producer to send the pending packet.
*/
-static inline bool hv_need_to_signal_on_read(struct hv_ring_buffer_info *rbi)
+static inline void hv_signal_on_read(struct vmbus_channel *channel)
{
u32 cur_write_sz;
u32 pending_sz;
+ struct hv_ring_buffer_info *rbi = &channel->inbound;
/*
* Issue a full memory barrier before making the signaling decision.
@@ -1501,14 +1502,14 @@ static inline bool hv_need_to_signal_on_read(struct hv_ring_buffer_info *rbi)
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 false;
+ return;
cur_write_sz = hv_get_bytes_to_write(rbi);
if (cur_write_sz >= pending_sz)
- return true;
+ vmbus_setevent(channel);
- return false;
+ return;
}
/*
@@ -1590,8 +1591,7 @@ static inline void commit_rd_index(struct vmbus_channel *channel)
virt_rmb();
ring_info->ring_buffer->read_index = ring_info->priv_read_index;
- if (hv_need_to_signal_on_read(ring_info))
- vmbus_set_event(channel);
+ hv_signal_on_read(channel);
}
struct vmpipe_proto_header {
--
2.11.0

140
alpine/kernel/patches-4.9/0011-Drivers-hv-vmbus-finally-fix-hv_need_to_signal_on_re.patch Normal file
View File

@ -0,0 +1,140 @@
From c9dc13e8f77c09369dae49f265176999e3f327c8 Mon Sep 17 00:00:00 2001
From: Dexuan Cui <decui@microsoft.com>
Date: Mon, 16 Jan 2017 15:46:19 +0800
Subject: [PATCH 11/11] Drivers: hv: vmbus: finally fix
hv_need_to_signal_on_read()
Commit a389fcfd2cb5 ("Drivers: hv: vmbus: Fix signaling logic in hv_need_to_signal_on_read()")
added the proper mb(), but removed the test "prev_write_sz < pending_sz"
when making the signal decision.
As a result, the guest can signal the host unnecessarily,
and then the host can throttle the guest because the host
thinks the guest is buggy or malicious; finally the user
running stress test can perceive intermittent freeze of
the guest.
This patch brings back the test, and properly handles the
in-place consumption APIs used by NetVSC (see get_next_pkt_raw(),
put_pkt_raw() and commit_rd_index()).
Fixes: a389fcfd2cb5 ("Drivers: hv: vmbus: Fix signaling logic in hv_need_to_signal_on_read()")
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Reported-by: Rolf Neugebauer <rolf.neugebauer@docker.com>
Tested-by: Rolf Neugebauer <rolf.neugebauer@docker.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: <stable@vger.kernel.org>
Origin: Email from Dexuan
---
drivers/hv/ring_buffer.c | 1 +
drivers/net/hyperv/netvsc.c | 2 ++
include/linux/hyperv.h | 32 ++++++++++++++++++++++++++++++--
3 files changed, 33 insertions(+), 2 deletions(-)
diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index cd49cb17eb7f..308dbda700eb 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c
@@ -383,6 +383,7 @@ int hv_ringbuffer_read(struct vmbus_channel *channel,
return ret;
}
+ init_cached_read_index(channel);
next_read_location = hv_get_next_read_location(inring_info);
next_read_location = hv_copyfrom_ringbuffer(inring_info, &desc,
sizeof(desc),
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 720b5fa9e625..9cd74817b09c 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -1288,6 +1288,8 @@ void netvsc_channel_cb(void *context)
ndev = hv_get_drvdata(device);
buffer = get_per_channel_state(channel);
+ init_cached_read_index(channel);
+
do {
desc = get_next_pkt_raw(channel);
if (desc != NULL) {
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index 5343b7406acc..e34da6846348 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -128,6 +128,7 @@ struct hv_ring_buffer_info {
u32 ring_data_startoffset;
u32 priv_write_index;
u32 priv_read_index;
+ u32 cached_read_index;
};
/*
@@ -180,6 +181,19 @@ static inline u32 hv_get_bytes_to_write(struct hv_ring_buffer_info *rbi)
return write;
}
+static inline u32 hv_get_cached_bytes_to_write(
+ const struct hv_ring_buffer_info *rbi)
+{
+ u32 read_loc, write_loc, dsize, write;
+
+ dsize = rbi->ring_datasize;
+ read_loc = rbi->cached_read_index;
+ write_loc = rbi->ring_buffer->write_index;
+
+ write = write_loc >= read_loc ? dsize - (write_loc - read_loc) :
+ read_loc - write_loc;
+ return write;
+}
/*
* VMBUS version is 32 bit entity broken up into
* two 16 bit quantities: major_number. minor_number.
@@ -1482,7 +1496,7 @@ hv_get_ring_buffer(struct hv_ring_buffer_info *ring_info)
static inline void hv_signal_on_read(struct vmbus_channel *channel)
{
- u32 cur_write_sz;
+ u32 cur_write_sz, cached_write_sz;
u32 pending_sz;
struct hv_ring_buffer_info *rbi = &channel->inbound;
@@ -1506,12 +1520,24 @@ static inline void hv_signal_on_read(struct vmbus_channel *channel)
cur_write_sz = hv_get_bytes_to_write(rbi);
- if (cur_write_sz >= pending_sz)
+ if (cur_write_sz < pending_sz)
+ return;
+
+ cached_write_sz = hv_get_cached_bytes_to_write(rbi);
+ if (cached_write_sz < pending_sz)
vmbus_setevent(channel);
return;
}
+static inline void
+init_cached_read_index(struct vmbus_channel *channel)
+{
+ struct hv_ring_buffer_info *rbi = &channel->inbound;
+
+ rbi->cached_read_index = rbi->ring_buffer->read_index;
+}
+
/*
* An API to support in-place processing of incoming VMBUS packets.
*/
@@ -1573,6 +1599,8 @@ static inline void put_pkt_raw(struct vmbus_channel *channel,
* This call commits the read index and potentially signals the host.
* Here is the pattern for using the "in-place" consumption APIs:
*
+ * init_cached_read_index();
+ *
* while (get_next_pkt_raw() {
* process the packet "in-place";
* put_pkt_raw();
--
2.11.0