13 Dec, 2018
1 commit
-
commit 37c2578c0c40e286bc0d30bdc05290b2058cf66e upstream.
vmbus_process_offer() mustn't call channel->sc_creation_callback()
directly for sub-channels, because sc_creation_callback() ->
vmbus_open() may never get the host's response to the
OPEN_CHANNEL message (the host may rescind a channel at any time,
e.g. in the case of hot removing a NIC), and vmbus_onoffer_rescind()
may not wake up the vmbus_open() as it's blocked due to a non-zero
vmbus_connection.offer_in_progress, and finally we have a deadlock.The above is also true for primary channels, if the related device
drivers use sync probing mode by default.And, usually the handling of primary channels and sub-channels can
depend on each other, so we should offload them to different
workqueues to avoid possible deadlock, e.g. in sync-probing mode,
NIC1's netvsc_subchan_work() can race with NIC2's netvsc_probe() ->
rtnl_lock(), and causes deadlock: the former gets the rtnl_lock
and waits for all the sub-channels to appear, but the latter
can't get the rtnl_lock and this blocks the handling of sub-channels.The patch can fix the multiple-NIC deadlock described above for
v3.x kernels (e.g. RHEL 7.x) which don't support async-probing
of devices, and v4.4, v4.9, v4.14 and v4.18 which support async-probing
but don't enable async-probing for Hyper-V drivers (yet).The patch can also fix the hang issue in sub-channel's handling described
above for all versions of kernels, including v4.19 and v4.20-rc4.So actually the patch should be applied to all the existing kernels,
not only the kernels that have 8195b1396ec8.Fixes: 8195b1396ec8 ("hv_netvsc: fix deadlock on hotplug")
Cc: stable@vger.kernel.org
Cc: Stephen Hemminger
Cc: K. Y. Srinivasan
Cc: Haiyang Zhang
Signed-off-by: Dexuan Cui
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
10 Oct, 2018
1 commit
-
commit 41e270f6898e7502be9fd6920ee0a108ca259d36 upstream.
With CONFIG_DEBUG_PREEMPT=y, I always see this warning:
BUG: using smp_processor_id() in preemptible [00000000]Fix the false warning by using get/put_cpu().
Here vmbus_connect() sends a message to the host and waits for the
host's response. The host will deliver the response message and an
interrupt on CPU msg->target_vcpu, and later the interrupt handler
will wake up vmbus_connect(). vmbus_connect() doesn't really have
to run on the same cpu as CPU msg->target_vcpu, so it's safe to
call put_cpu() just here.Signed-off-by: Dexuan Cui
Cc: stable@vger.kernel.org
Cc: K. Y. Srinivasan
Cc: Haiyang Zhang
Cc: Stephen Hemminger
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
10 Aug, 2017
3 commits
-
To support implementing remote TLB flushing on Hyper-V with a hypercall
we need to make vp_index available outside of vmbus module. Rename and
globalize.Signed-off-by: Vitaly Kuznetsov
Reviewed-by: Andy Shevchenko
Reviewed-by: Stephen Hemminger
Cc: Andy Lutomirski
Cc: Haiyang Zhang
Cc: Jork Loeser
Cc: K. Y. Srinivasan
Cc: Linus Torvalds
Cc: Peter Zijlstra
Cc: Simon Xiao
Cc: Steven Rostedt
Cc: Thomas Gleixner
Cc: devel@linuxdriverproject.org
Link: http://lkml.kernel.org/r/20170802160921.21791-7-vkuznets@redhat.com
Signed-off-by: Ingo Molnar -
We need to pass only 8 bytes of input for HvSignalEvent which makes it a
perfect fit for fast hypercall. hv_input_signal_event_buffer is not needed
any more and hv_input_signal_event is converted to union for convenience.Signed-off-by: Vitaly Kuznetsov
Reviewed-by: Andy Shevchenko
Reviewed-by: Stephen Hemminger
Cc: Andy Lutomirski
Cc: Haiyang Zhang
Cc: Jork Loeser
Cc: K. Y. Srinivasan
Cc: Linus Torvalds
Cc: Peter Zijlstra
Cc: Simon Xiao
Cc: Steven Rostedt
Cc: Thomas Gleixner
Cc: devel@linuxdriverproject.org
Link: http://lkml.kernel.org/r/20170802160921.21791-5-vkuznets@redhat.com
Signed-off-by: Ingo Molnar -
We have only three call sites for hv_do_hypercall() and we're going to
change HVCALL_SIGNAL_EVENT to doing fast hypercall so we can inline this
function for optimization.Hyper-V top level functional specification states that r9-r11 registers
and flags may be clobbered by the hypervisor during hypercall and with
inlining this is somewhat important, add the clobbers.Signed-off-by: Vitaly Kuznetsov
Reviewed-by: Andy Shevchenko
Reviewed-by: Stephen Hemminger
Cc: Andy Lutomirski
Cc: Haiyang Zhang
Cc: Jork Loeser
Cc: K. Y. Srinivasan
Cc: Linus Torvalds
Cc: Peter Zijlstra
Cc: Simon Xiao
Cc: Steven Rostedt
Cc: Thomas Gleixner
Cc: devel@linuxdriverproject.org
Link: http://lkml.kernel.org/r/20170802160921.21791-3-vkuznets@redhat.com
Signed-off-by: Ingo Molnar
25 May, 2017
1 commit
-
Commit c0bb03924f1a ("Drivers: hv: vmbus: Raise retry/wait limits in
vmbus_post_msg()") increased the retry/wait limits of vmbus_post_msg()
to address the new DoS protection policies in WS2016.
Increase the time between retries to make the
code more robust.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
18 May, 2017
2 commits
-
Fix the rescind handling. This patch addresses the following rescind
scenario that is currently not handled correctly:If a rescind were to be received while the offer is still being
peocessed, we will be blocked indefinitely since the rescind message
is handled on the same work element as the offer message. Fix this
issue.I would like to thank Dexuan Cui and
Long Li for working with me on this patch.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman -
ENOBUFS is a more approrpiate error code to be returned
when the hypervisor cannot post the message because of
insufficient buffers. Make the adjustment.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
17 Mar, 2017
1 commit
-
The change to reschedule tasklet if more data arrives in ring buffer
can cause performance regression if host timing is such that the
next response happens in small window.Go back to a modified version of the original looping behavior.
If the race occurs in a small time, then loop. But if the tasklet
has been running for a long interval due to flood, then reschedule
the tasklet to allow migration to ksoftirqd.Signed-off-by: Stephen Hemminger
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
15 Feb, 2017
3 commits
-
Change the simple boolean batched_reading into a tri-value.
For future NAPI support in netvsc driver, the callback needs to
occur directly in interrupt handler.Batched mode is also changed to disable host interrupts immediately
in interrupt routine (to avoid unnecessary host signals), and the
tasklet is rescheduled if more data is detected.Signed-off-by: Stephen Hemminger
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman -
Make the event handling tasklet per channel rather than per-cpu.
This allows for better fairness when getting lots of data on the same
cpu.Signed-off-by: Stephen Hemminger
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman -
The hv_context structure had several arrays which were per-cpu
and was allocating small structures (tasklet_struct). Instead use
a single per-cpu array.Signed-off-by: Stephen Hemminger
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
10 Feb, 2017
1 commit
-
Use standard kernel operations for find first set bit to traverse
the channel bit array. This has added benefit of speeding up
lookup on 64 bit and because it uses find first set instruction.Signed-off-by: Stephen Hemminger
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
20 Jan, 2017
1 commit
-
As part of the effort to separate out architecture specific code,
extract hypervisor version information in an architecture specific
file.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
11 Jan, 2017
1 commit
-
DoS protection conditions were altered in WS2016 and now it's easy to get
-EAGAIN returned from vmbus_post_msg() (e.g. when we try changing MTU on a
netvsc device in a loop). All vmbus_post_msg() callers don't retry the
operation and we usually end up with a non-functional device or crash.While host's DoS protection conditions are unknown to me my tests show that
it can take up to 10 seconds before the message is sent so doing udelay()
is not an option, we really need to sleep. Almost all vmbus_post_msg()
callers are ready to sleep but there is one special case:
vmbus_initiate_unload() which can be called from interrupt/NMI context and
we can't sleep there. I'm also not sure about the lonely
vmbus_send_tl_connect_request() which has no in-tree users but its external
users are most likely waiting for the host to reply so sleeping there is
also appropriate.Signed-off-by: Vitaly Kuznetsov
Signed-off-by: K. Y. Srinivasan
Cc:
Signed-off-by: Greg Kroah-Hartman
06 Dec, 2016
1 commit
-
This is a new driver to enable userspace networking on VMBus.
It is based largely on the similar driver that already exists
for PCI, and earlier work done by Brocade to support DPDK.Signed-off-by: Stephen Hemminger
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
31 Aug, 2016
1 commit
-
The current delay between retries is unnecessarily high and is negatively
affecting the time it takes to boot the system.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
01 May, 2016
1 commit
-
In preparation for moving some ring buffer functionality out of the
vmbus driver, export the API for signaling the host.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
02 Mar, 2016
2 commits
-
WS2012 R2 and above hosts can support kexec in that thay can support
reconnecting to the host (as would be needed in the kexec path)
on any CPU. Enable this. Pre ws2012 r2 hosts don't have this ability
and consequently cannot support kexec.Signed-off-by: Alex Ng
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman -
wait_for_completion() may sleep, it enables interrupts and this
is something we really want to avoid on crashes because interrupt
handlers can cause other crashes. Switch to the recently introduced
vmbus_wait_for_unload() doing busy wait instead.Reported-by: Radim Krcmar
Signed-off-by: Vitaly Kuznetsov
Reviewed-by: Radim Kr.má
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
08 Feb, 2016
2 commits
-
The hvsock driver needs this API to release all the resources related
to the channel.Signed-off-by: Dexuan Cui
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman -
Cleanup vmbus_set_event() by inlining the hypercall to post
the event and since the return value of vmbus_set_event() is not checked,
make it void. As part of this cleanup, get rid of the function
hv_signal_event() as it is only callled from vmbus_set_event().Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
15 Dec, 2015
2 commits
-
Force all channel messages to be delivered on CPU0. These messages are not
performance critical and are used during the setup and teardown of the
channel.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman -
spinlock is unnecessary here.
mutex is enough.Signed-off-by: Dexuan Cui
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
01 Jun, 2015
1 commit
-
Add support for Windows 10.
Signed-off-by: Keith Mange
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
25 May, 2015
1 commit
-
Implement the protocol for tearing down the monitor state established with
the host.Signed-off-by: K. Y. Srinivasan
Tested-by: Vitaly Kuznetsov
Signed-off-by: Greg Kroah-Hartman
03 Apr, 2015
2 commits
-
Most of the retries can be done within a millisecond successfully, so we
sleep 1ms before the first retry, then gradually increase the retry
interval to 2^n with max value of 2048ms. Doing so, we will have shorter
overall delay time, because most of the cases succeed within 1-2 attempts.Signed-off-by: Haiyang Zhang
Reviewed-by: K. Y. Srinivasan
Reviewed-by: Dexuan Cui
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman -
Since the 2 fucntions can safely run in vmbus_connection.work_queue without
hang, we don't need to schedule new work items into the per-channel workqueue.Actally we can even remove the per-channel workqueue now -- we'll do it
in the next patch.Signed-off-by: Dexuan Cui
Cc: K. Y. Srinivasan
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
25 Mar, 2015
1 commit
-
This patch is a continuation of the rescind handling cleanup work. We cannot
block in the global message handling work context especially if we are blocking
waiting for the host to wake us up. I would like to thank
Dexuan Cui for observing this problem.The current char-next branch is broken and this patch fixes
the bug.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
02 Mar, 2015
3 commits
-
Currently we log messages when either we are not able to map an ID to a
channel or when the channel does not have a callback associated
(in the channel interrupt handling path). These messages don't add
any value, get rid of them.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman -
I got HV_STATUS_INVALID_CONNECTION_ID on Hyper-V 2008 R2 when keeping running
"rmmod hv_netvsc; modprobe hv_netvsc; rmmod hv_utils; modprobe hv_utils"
in a Linux guest. Looks the host has some kind of throttling mechanism if
some kinds of hypercalls are sent too frequently.
Without the patch, the driver can occasionally fail to load.Also let's retry HV_STATUS_INSUFFICIENT_MEMORY, though we didn't get it
before.Removed 'case -ENOMEM', since the hypervisor doesn't return this.
CC: "K. Y. Srinivasan"
Reviewed-by: Jason Wang
Signed-off-by: Dexuan Cui
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman -
We need to destroy hv_vmbus_con on module shutdown, otherwise the following
crash is sometimes observed:[ 76.569845] hv_vmbus: Hyper-V Host Build:9600-6.3-17-0.17039; Vmbus version:3.0
[ 82.598859] BUG: unable to handle kernel paging request at ffffffffa0003480
[ 82.599287] IP: [] 0xffffffffa0003480
[ 82.599287] PGD 1f34067 PUD 1f35063 PMD 3f72d067 PTE 0
[ 82.599287] Oops: 0010 [#1] SMP
[ 82.599287] Modules linked in: [last unloaded: hv_vmbus]
[ 82.599287] CPU: 0 PID: 26 Comm: kworker/0:1 Not tainted 3.19.0-rc5_bug923184+ #488
[ 82.599287] Hardware name: Microsoft Corporation Virtual Machine/Virtual Machine, BIOS Hyper-V UEFI Release v1.0 11/26/2012
[ 82.599287] Workqueue: hv_vmbus_con 0xffffffffa0003480
[ 82.599287] task: ffff88007b6ddfa0 ti: ffff88007f8f8000 task.ti: ffff88007f8f8000
[ 82.599287] RIP: 0010:[] [] 0xffffffffa0003480
[ 82.599287] RSP: 0018:ffff88007f8fbe00 EFLAGS: 00010202
...To avoid memory leaks we need to free monitor_pages and int_page for
vmbus_connection. Implement vmbus_disconnect() function by separating cleanup
path from vmbus_connect().As we use hv_vmbus_con to release channels (see free_channel() in channel_mgmt.c)
we need to make sure the work was done before we remove the queue, do that with
drain_workqueue(). We also need to avoid handling messages which can (potentially)
create new channels, so set vmbus_connection.conn_state = DISCONNECTED at the very
beginning of vmbus_exit() and check for that in vmbus_onmessage_work().Signed-off-by: Vitaly Kuznetsov
Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
12 Jan, 2015
1 commit
-
Replace calls for smp_processor_id() to get_cpu() to get the CPU ID of
the current CPU. In these instances, there is no correctness issue with
regards to preemption, we just need the current CPU ID.Signed-off-by: K. Y. Srinivasan
Signed-off-by: Greg Kroah-Hartman
24 Sep, 2014
1 commit
-
Posting messages to the host can fail because of transient resource
related failures. Correctly deal with these failures and increase the
number of attempts to post the message before giving up.In this version of the patch, I have normalized the error code to
Linux error code.Signed-off-by: K. Y. Srinivasan
Cc:
Tested-by: Sitsofe Wheeler
Signed-off-by: Greg Kroah-Hartman
10 Jul, 2014
1 commit
-
Starting with Win8, we have implemented several optimizations to improve the
scalability and performance of the VMBUS transport between the Host and the
Guest. Some of the non-performance critical services cannot leverage these
optimization since they only read and process one message at a time.
Make adjustments to the callback dispatch code to account for the way
non-performance critical drivers handle reading of the channel.Signed-off-by: K. Y. Srinivasan
Cc:
Signed-off-by: Greg Kroah-Hartman
29 May, 2014
1 commit
-
We try to free two pages when only one has been allocated.
Cleanup path is unlikely, so I haven't found any trace that would fit,
but I hope that free_pages_prepare() does catch it.Cc: stable@vger.kernel.org
Signed-off-by: Radim Krčmář
Reviewed-by: Amos Kong
Acked-by: Jason Wang
Signed-off-by: Greg Kroah-Hartman
04 May, 2014
2 commits
-
Currently the mapping of the relID to channel is done under the protection of a
single spin lock. Starting with ws2012, each channel is bound to a specific VCPU
in the guest. Use this binding to eliminate the spin lock by setting up
per-cpu state for mapping relId to the channel.Signed-off-by: K. Y. Srinivasan
Reviewed-by: Haiyang Zhang
Signed-off-by: Greg Kroah-Hartman -
By ensuring that we set the callback handler to NULL in the channel close
path on the same CPU that the channel is bound to, we can eliminate this lock
acquisition and release in a performance critical path.Signed-off-by: K. Y. Srinivasan
Reviewed-by: Haiyang Zhang
Signed-off-by: Greg Kroah-Hartman
17 Apr, 2014
1 commit
-
Only ws2012r2 hosts support the ability to reconnect to the host on VMBUS. This functionality
is needed by kexec in Linux. To use this functionality we need to negotiate version 3.0 of the
VMBUS protocol.Signed-off-by: K. Y. Srinivasan
Cc: [3.9+]
Signed-off-by: Greg Kroah-Hartman
08 Feb, 2014
1 commit
-
When the guest attempts to connect with the host when there may already be a
connection with the host (as would be the case during the kdump/kexec path),
it is difficult to guarantee timely response from the host. Starting with
WS2012 R2, the host supports this ability to re-connect with the host
(explicitly to support kexec). Prior to responding to the guest, the host
needs to ensure that device states based on the previous connection to
the host have been properly torn down. This may introduce unbounded delays.
To deal with this issue, don't do a timed wait during the initial connect
with the host.Signed-off-by: K. Y. Srinivasan
Cc:
Signed-off-by: Greg Kroah-Hartman