Sasha writes:
hyperv-fixes-4.20-rc6
* tag 'hyperv-fixes-4.20-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux:
Drivers: hv: vmbus: Return -EINVAL for the sys files for unopened channels
x86, hyperv: remove PCI dependency
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 8195b1396e.
Fixes: 8195b1396e ("hv_netvsc: fix deadlock on hotplug")
Cc: stable@vger.kernel.org
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: K. Y. Srinivasan <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This is a longstanding issue: if the vmbus upper-layer drivers try to
consume too many GPADLs, the host may return with an error
0xC0000044 (STATUS_QUOTA_EXCEEDED), but currently we forget to check
the creation_status, and hence we can pass an invalid GPADL handle
into the OPEN_CHANNEL message, and get an error code 0xc0000225 in
open_info->response.open_result.status, and finally we hang in
vmbus_open() -> "goto error_free_info" -> vmbus_teardown_gpadl().
With this patch, we can exit gracefully on STATUS_QUOTA_EXCEEDED.
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: K. Y. Srinivasan <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In kvp_send_key(), we do need call process_ib_ipinfo() if
message->kvp_hdr.operation is KVP_OP_GET_IP_INFO, because it turns out
the userland hv_kvp_daemon needs the info of operation, adapter_id and
addr_family. With the incorrect fc62c3b197, the host can't get the
VM's IP via KVP.
And, fc62c3b197 added a "break;", but actually forgot to initialize
the key_size/value in the case of KVP_OP_SET, so the default key_size of
0 is passed to the kvp daemon, and the pool files
/var/lib/hyperv/.kvp_pool_* can't be updated.
This patch effectively rolls back the previous fc62c3b197, and
correctly fixes the "this statement may fall through" warnings.
This patch is tested on WS 2012 R2 and 2016.
Fixes: fc62c3b197 ("Drivers: hv: kvp: Fix two "this statement may fall through" warnings")
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Cc: K. Y. Srinivasan <kys@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Cc: <Stable@vger.kernel.org>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
lockdep_assert_held() is better suited to checking locking requirements,
since it won't get confused when someone else holds the lock. This is
also a step towards possibly removing spin_is_locked().
Signed-off-by: Lance Roy <ldr709@gmail.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A cpumask structure on the stack can cause a warning with
CONFIG_NR_CPUS=8192 (e.g. Ubuntu 16.04 and 18.04 use this):
drivers/hv//channel_mgmt.c: In function ‘init_vp_index’:
drivers/hv//channel_mgmt.c:702:1: warning: the frame size of 1032 bytes
is larger than 1024 bytes [-Wframe-larger-than=]
Nowadays it looks most distros enable CONFIG_CPUMASK_OFFSTACK=y, and
hence we can work around the warning by using cpumask_var_t.
Signed-off-by: Dexuan Cui <decui@microsoft.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>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We don't need to call process_ib_ipinfo() if message->kvp_hdr.operation is
KVP_OP_GET_IP_INFO in kvp_send_key(), because here we just need to pass on
the op code from the host to the userspace; when the userspace returns
the info requested by the host, we pass the info on to the host in
kvp_respond_to_host() -> process_ob_ipinfo(). BTW, the current buggy code
actually doesn't cause any harm, because only message->kvp_hdr.operation
is used by the userspace, in the case of KVP_OP_GET_IP_INFO.
The patch also adds a missing "break;" in kvp_send_key(). BTW, the current
buggy code actually doesn't cause any harm, because in the case of
KVP_OP_SET, the unexpected fall-through corrupts
message->body.kvp_set.data.key_size, but that is not really used: see
the definition of struct hv_kvp_exchg_msg_value.
Signed-off-by: Dexuan Cui <decui@microsoft.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>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
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 <decui@microsoft.com>
Cc: stable@vger.kernel.org
Cc: K. Y. Srinivasan <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The UIO driver needs the ring buffer to be persistent(reused)
across open/close. Split the allocation and setup of ring buffer
out of vmbus_open. For normal usage vmbus_open/vmbus_close there
are no changes; only impacts uio_hv_generic which needs to keep
ring buffer memory and reuse when application restarts.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Avoid going from struct page to virt address (and back) by just
keeping pointer to the allocated pages instead of virt address.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Rather than passing relid and then looking up the channel.
Pass the channel directly, since caller already knows it.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If hv_synic_alloc() errors out, the state of the per-cpu context
for some CPUs is unknown since the zero'ing is done as each
CPU is iterated over. In such case, hv_synic_cleanup() may try to
free memory based on uninitialized values. Fix this by zero'ing
the per-cpu context for all CPUs before doing any memory
allocations that might fail.
Signed-off-by: Michael Kelley <mikelley@microsoft.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For unsupported device types, the vmbus channel ringbuffer is never
initialized, and therefore reading the sysfs files will return garbage
or cause a kernel OOPS.
Fixes: c2e5df616e ("vmbus: add per-channel sysfs info")
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Cc: <stable@vger.kernel.org> # 4.15
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
clk_evt memory is not being freed when the synic is shutdown
or when there is an allocation error. Add the appropriate
kfree() call, along with a comment to clarify how the memory
gets freed after an allocation error. Make the free path
consistent by removing checks for NULL since kfree() and
free_page() already do the check.
Signed-off-by: Michael Kelley <mikelley@microsoft.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
slow_virt_to_phys() is only implemented for arch/x86.
Remove its use in arch independent Hyper-V drivers, and
replace with test for vmalloc() address followed by
appropriate v-to-p function. This follows the typical
pattern of other drivers and avoids the need to implement
slow_virt_to_phys() for Hyper-V on ARM64.
Signed-off-by: Michael Kelley <mikelley@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Before setting channel->rescind in vmbus_rescind_cleanup(), we should make
sure the channel callback won't run any more, otherwise a high-level
driver like pci_hyperv, which may be infinitely waiting for the host VSP's
response and notices the channel has been rescinded, can't safely give
up: e.g., in hv_pci_protocol_negotiation() -> wait_for_response(), it's
unsafe to exit from wait_for_response() and proceed with the on-stack
variable "comp_pkt" popped. The issue was originally spotted by
Michael Kelley <mikelley@microsoft.com>.
In vmbus_close_internal(), the patch also minimizes the range protected by
disabling/enabling channel->callback_event: we don't really need that for
the whole function.
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Cc: stable@vger.kernel.org
Cc: K. Y. Srinivasan <kys@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Michael Kelley <mikelley@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Being able to find the numa_node for a device is useful for userspace
drivers (DPDK) and also for diagnosing performance issues. This makes
vmbus similar to pci.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Get rid of ISA specific code from vmus_drv.c which is common code.
Fixes: 81b18bce48 ("Drivers: HV: Send one page worth of kmsg dump over Hyper-V during panic")
Signed-off-by: Sunil Muthuswamy <sunilmut@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Greg Kroah-Hartman
Dexuan Cui
Sinan Kaya
Lance Roy
Stephen Hemminger
Michael Kelley
Sunil Muthuswamy