- 29 Feb, 2016 40 commits
-
-
Manoj Kumar authored
After a few iterations of resetting the card, either during EEH recovery, or a host_reset the following is seen in the logs. cxlflash 0008:00: cxlflash_queuecommand: could not get a free command At every reset of the card, the commands that are outstanding are being leaked. No effort is being made to reap these commands. A few more resets later, the above error message floods the logs and the card is rendered totally unusable as no free commands are available. Iterated through the 'cmd' queue and printed out the 'free' counter and found that on each reset certain commands were in-use and stayed in-use through subsequent resets. To resolve this issue, when the card is reset, reap all the commands that are active/outstanding. Signed-off-by:
Manoj N. Kumar <manoj@linux.vnet.ibm.com> Acked-by:
Matthew R. Ochs <mrochs@linux.vnet.ibm.com> Reviewed-by:
Andrew Donnellan <andrew.donnellan@au1.ibm.com> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com> (cherry picked from commit ee91e332) Signed-off-by:
Tim Gardner <tim.gardner@canonical.com>
-
Uma Krishnan authored
Having a date for the driver requires it to be updated quite often. Removing the date which is not necessary. Also made use of the existing symbol to print the driver name. Signed-off-by:
Uma Krishnan <ukrishn@linux.vnet.ibm.com> Reviewed-by:
-
Vaibhav Jain authored
Presently when a user-space process issues CXL_IOCTL_START_WORK ioctl we store the pid of the current task_struct and use it to get pointer to the mm_struct of the process, while processing page or segment faults from the capi card. However this causes issues when the thread that had originally issued the start-work ioctl exits in which case the stored pid is no more valid and the cxl driver is unable to handle faults as the mm_struct corresponding to process is no more accessible. This patch fixes this issue by using the mm_struct of the next alive task in the thread group. This is done by iterating over all the tasks in the thread group starting from thread group leader and calling get_task_mm on each one of them. When a valid mm_struct is obtained the pid of the associated task is stored in the context replacing the exiting one for handling future faults. The patch introduces a new function named get_mem_context that checks if the current task pointed to by ctx->pid is dead? If yes it performs the steps described above. Also a new variable cxl_context.glpid is introduced which stores the pid of the thread group leader associated with the context owning task. Reported-by:
Frank Haverkamp <HAVERKAM@de.ibm.com> Suggested-by:
Ian Munsie <imunsie@au1.ibm.com> Signed-off-by:
Vaibhav Jain <vaibhav@linux.vnet.ibm.com> Acked-by:
Frederic Barrat <fbarrat@linux.vnet.ibm.com> Reviewed-by:
Michael Ellerman <mpe@ellerman.id.au> (cherry picked from commit 7b8ad495) Signed-off-by:
-
Geliang Tang authored
IS_ERR_OR_NULL already contain an unlikely compiler flag. Drop it. Signed-off-by:
Geliang Tang <geliangtang@163.com> Acked-by:
-
Andrew Donnellan authored
When writing a value to config space, cxl_pcie_write_config() calls cxl_pcie_config_info() to obtain a mask and shift value, shifts the new value accordingly, then uses the mask to combine the shifted value with the existing value at the address as part of a read-modify-write pattern. Currently, we use a logical OR operator rather than a bitwise OR operator, which means any use of this function results in an incorrect value being written. Replace the logical OR operator with a bitwise OR operator so the value is written correctly. Reported-by:
-
Vaibhav Jain authored
An idr warning is reported when a context is release after the capi card is unbound from the cxl driver via sysfs. Below are the steps to reproduce: 1. Create multiple afu contexts in an user-space application using libcxl. 2. Unbind capi card from cxl using command of form echo <capi-card-pci-addr> > /sys/bus/pci/drivers/cxl-pci/unbind 3. Exit/kill the application owning afu contexts. After above steps a warning message is usually seen in the kernel logs of the form "idr_remove called for id=<context-id> which is not allocated." This is caused by the function cxl_release_afu which destroys the contexts_idr table. So when a context is release no entry for context pe is found in the contexts_idr table and idr code prints this warning. This patch fixes this issue by increasing & decreasing the ref-count on the afu device when a context is initialized or when its freed respectively. This prevents the afu from being released until all the afu contexts have been released. The patch introduces two new functions namely cxl_afu_get/put that manage the ref-count on the afu device. Also the patch removes code inside cxl_dev_context_init that increases ref on the afu device as its guaranteed to be alive during this function. Reported-by:
-
Tim Gardner authored
Ignore: yes Signed-off-by: -
Tim Gardner authored
Signed-off-by: -
Tim Gardner authored
Signed-off-by: -
K. Y. Srinivasan authored
For performance critical devices, we distribute the incoming channel interrupt load across available CPUs in the guest. Include Fibre channel devices in the set of devices for which we would distribute the interrupt load. Signed-off-by:
K. Y. Srinivasan <kys@microsoft.com> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org> (cherry picked from commit 879a650a) Signed-off-by:
-
Vitaly Kuznetsov authored
The return type of hvt_op_poll() is unsigned int and -EBADF is inappropriate, poll functions return POLL* statuses. Reported-by:
Dexuan Cui <decui@microsoft.com> Signed-off-by:
Vitaly Kuznetsov <vkuznets@redhat.com> Signed-off-by:
-
Dexuan Cui authored
With the recent change af3ff643 (Drivers: hv: vmbus: Use uuid_le type consistently), we always get this warning: CC [M] drivers/input/serio/hyperv-keyboard.o drivers/input/serio/hyperv-keyboard.c:427:2: warning: missing braces around initializer [-Wmissing-braces] { HV_KBD_GUID, }, ^ drivers/input/serio/hyperv-keyboard.c:427:2: warning: (near initialization for .id_table[0].guid.b.) [-Wmissing-braces] The patch fixes the warning. Signed-off-by:
-
Andrey Smetanin authored
This struct is required for Hyper-V SynIC timers implementation inside KVM and for upcoming Hyper-V VMBus support by userspace(QEMU). So place it into Hyper-V UAPI header. Signed-off-by:
Andrey Smetanin <asmetanin@virtuozzo.com> CC: Gleb Natapov <gleb@kernel.org> CC: Paolo Bonzini <pbonzini@redhat.com> CC: "K. Y. Srinivasan" <kys@microsoft.com> CC: Haiyang Zhang <haiyangz@microsoft.com> CC: Vitaly Kuznetsov <vkuznets@redhat.com> CC: Roman Kagan <rkagan@virtuozzo.com> CC: Denis V. Lunev <den@openvz.org> CC: qemu-devel@nongnu.org Signed-off-by:
Paolo Bonzini <pbonzini@redhat.com> (cherry picked from commit c71acc4c) Signed-off-by:
-
Andrey Smetanin authored
This struct is required for Hyper-V SynIC timers implementation inside KVM and for upcoming Hyper-V VMBus support by userspace(QEMU). So place it into Hyper-V UAPI header. Signed-off-by:
Roman Kagan <rkagan@virtuozzo.com> CC: Gleb Natapov <gleb@kernel.org> CC: Paolo Bonzini <pbonzini@redhat.com> CC: "K. Y. Srinivasan" <kys@microsoft.com> CC: Haiyang Zhang <haiyangz@microsoft.com> CC: Vitaly Kuznetsov <vkuznets@redhat.com> CC: Roman Kagan <rkagan@virtuozzo.com> CC: Denis V. Lunev <den@openvz.org> CC: qemu-devel@nongnu.org Signed-off-by:
-
Andrey Smetanin authored
This constant is required for Hyper-V SynIC timers MSR's support by userspace(QEMU). Signed-off-by:
-
Andrey Smetanin authored
enum hv_message_type inside struct hv_message, hv_post_message is not size portable. Replace enum by u32. Signed-off-by:
-
Vitaly Kuznetsov authored
Currently, there is only one user for hv_ringbuffer_read()/ hv_ringbuffer_peak() functions and the usage of these functions is: - insecure as we drop ring_lock between them, someone else (in theory only) can acquire it in between; - non-optimal as we do a number of things (acquire/release the above mentioned lock, calculate available space on the ring, ...) twice and this path is performance-critical. Remove hv_ringbuffer_peek() moving the logic from __vmbus_recvpacket() to hv_ringbuffer_read(). Signed-off-by:
-
Vitaly Kuznetsov authored
vmbus_recvpacket() and vmbus_recvpacket_raw() are almost identical but there are two discrepancies: 1) vmbus_recvpacket() doesn't propagate errors from hv_ringbuffer_read() which looks like it is not desired. 2) There is an error message printed in packetlen > bufferlen case in vmbus_recvpacket(). I'm removing it as it is usless for users to see such messages and /vmbus_recvpacket_raw() doesn't have it. Signed-off-by:
-
Vitaly Kuznetsov authored
hv_ringbuffer_peek() does the same as hv_ringbuffer_read() without advancing the read index. The only functional change this patch brings is moving hv_need_to_signal_on_read() call under the ring_lock but this function is just a couple of comparisons. Signed-off-by:
-
Vitaly Kuznetsov authored
smp_read_barrier_depends() does nothing on almost all arcitectures including x86 and having it in the beginning of hv_get_ringbuffer_availbytes() does not provide any guarantees anyway. Signed-off-by:
-
Vitaly Kuznetsov authored
Convert 6+-string comments repeating function names to normal kernel-style comments and fix a couple of other comment style issues. No textual or functional changes intended. Signed-off-by:
-
Vitaly Kuznetsov authored
The crash is observed when a service is being disabled host side while userspace daemon is connected to the device: [ 90.244859] general protection fault: 0000 [#1] SMP ... [ 90.800082] Call Trace: [ 90.800082] [<ffffffff81187008>] __fput+0xc8/0x1f0 [ 90.800082] [<ffffffff8118716e>] ____fput+0xe/0x10 ... [ 90.800082] [<ffffffff81015278>] do_signal+0x28/0x580 [ 90.800082] [<ffffffff81086656>] ? finish_task_switch+0xa6/0x180 [ 90.800082] [<ffffffff81443ebf>] ? __schedule+0x28f/0x870 [ 90.800082] [<ffffffffa01ebbaa>] ? hvt_op_read+0x12a/0x140 [hv_utils] ... The problem is that hvutil_transport_destroy() which does misc_deregister() freeing the appropriate device is reachable by two paths: module unload and from util_remove(). While module unload path is protected by .owner in struct file_operations util_remove() path is not. Freeing the device while someone holds an open fd for it is a show stopper. In general, it is not possible to revoke an fd from all users so the only way to solve the issue is to defer freeing the hvutil_transport structure. Signed-off-by:
-
Vitaly Kuznetsov authored
When Hyper-V host asks us to remove some util driver by closing the appropriate channel there is no easy way to force the current file descriptor holder to hang up but we can start to respond -EBADF to all operations asking it to exit gracefully. As we're setting hvt->mode from two separate contexts now we need to use a proper locking. Signed-off-by:
-
Vitaly Kuznetsov authored
As a preparation to reusing outmsg_lock to protect test-and-set openrations on 'mode' rename it the more general 'lock'. Signed-off-by:
-
Vitaly Kuznetsov authored
inmsg should be freed in case of on_msg() failure to avoid memory leak. Preserve the error code from on_msg(). Signed-off-by:
-
Dexuan Cui authored
Fix the write()'s argument in the daemon code. Cc: Vitaly Kuznetsov <vkuznets@redhat.com> Cc: "K. Y. Srinivasan" <kys@microsoft.com> Signed-off-by:
-
K. Y. Srinivasan authored
When the handshake with daemon is complete, we should poll the channel since during the handshake, we will not be processing any messages. This is a potential bug if the host is waiting for a response from the guest. I would like to thank Dexuan for pointing this out. Signed-off-by:
-
K. Y. Srinivasan authored
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:
-
Andrey Smetanin authored
Hypervisor Top Level Functional Specification v3/4 says that TSC page sequence value = -1(0xFFFFFFFF) is used to indicate that TSC page no longer reliable source of reference timer. Unfortunately, we found that Windows Hyper-V guest side implementation uses sequence value = 0 to indicate that Tsc page no longer valid. This is clearly visible inside Windows 2012R2 ntoskrnl.exe HvlGetReferenceTime() function dissassembly: HvlGetReferenceTime proc near xchg ax, ax loc_1401C3132: mov rax, cs:HvlpReferenceTscPage mov r9d, [rax] test r9d, r9d jz short loc_1401C3176 rdtsc mov rcx, cs:HvlpReferenceTscPage shl rdx, 20h or rdx, rax mov rax, [rcx+8] mov rcx, cs:HvlpReferenceTscPage mov r8, [rcx+10h] mul rdx mov rax, cs:HvlpReferenceTscPage add rdx, r8 mov ecx, [rax] cmp ecx, r9d jnz short loc_1401C3132 jmp short loc_1401C3184 loc_1401C3176: mov ecx, 40000020h rdmsr shl rdx, 20h or rdx, rax loc_1401C3184: mov rax, rdx retn HvlGetReferenceTime endp This patch aligns Tsc page invalid sequence value with Windows Hyper-V guest implementation which is more compatible with both Hyper-V hypervisor and KVM hypervisor. Signed-off-by:Denis V. Lunev <den@openvz.org> CC: "K. Y. Srinivasan" <kys@microsoft.com> CC: Haiyang Zhang <haiyangz@microsoft.com> CC: Vitaly Kuznetsov <vkuznets@redhat.com> Signed-off-by:
-
K. Y. Srinivasan authored
Currently we have two policies for deciding when to signal the host: One based on the ring buffer state and the other based on what the VMBUS client driver wants to do. Consider the case when the client wants to explicitly control when to signal the host. In this case, if the client were to defer signaling, we will not be able to signal the host subsequently when the client does want to signal since the ring buffer state will prevent the signaling. Implement logic to have only one signaling policy in force for a given channel. Signed-off-by:
Haiyang Zhang <haiyangz@microsoft.com> Tested-by:
-
Jake Oshins authored
This patch makes 16GB GPUs work in Hyper-V VMs, since, for compatibility reasons, the Hyper-V BIOS lists MMIO ranges in 2GB chunks in its root bus's _CRS object. Signed-off-by:
Jake Oshins <jakeo@microsoft.com> Signed-off-by:
-
Dexuan Cui authored
spinlock is unnecessary here. mutex is enough. Signed-off-by:
-
Dexuan Cui authored
We want to simplify vmbus_onoffer_rescind() by not invoking hv_process_channel_removal(NULL, ...). Signed-off-by:
-
Dexuan Cui authored
In the path vmbus_onoffer_rescind() -> vmbus_device_unregister() -> device_unregister() -> ... -> __device_release_driver(), we can see for a device without a driver loaded: dev->driver is NULL, so dev->bus->remove(dev), namely vmbus_remove(), isn't invoked. As a result, vmbus_remove() -> hv_process_channel_removal() isn't invoked and some cleanups(like sending a CHANNELMSG_RELID_RELEASED message to the host) aren't done. We can demo the issue this way: 1. rmmod hv_utils; 2. disable the Heartbeat Integration Service in Hyper-V Manager and lsvmbus shows the device disappears. 3. re-enable the Heartbeat in Hyper-V Manager and modprobe hv_utils, but lsvmbus shows the device can't appear again. This is because, the host thinks the VM hasn't released the relid, so can't re-offer the device to the VM. We can fix the issue by moving hv_process_channel_removal() from vmbus_close_internal() to vmbus_device_release(), since the latter is always invoked on device_unregister(), whether or not the dev has a driver loaded. Signed-off-by:
-
Dexuan Cui authored
This fixes an incorrect assumption of channel state in the function. Signed-off-by:
-
Dexuan Cui authored
process_chn_event(), running in the tasklet, can race with vmbus_close_internal() in the case of SMP guest, e.g., when the former is accessing channel->inbound.ring_buffer, the latter could be freeing the ring_buffer pages. To resolve the race, we can serialize them by disabling the tasklet when the latter is running here. Signed-off-by:
-
K. Y. Srinivasan authored
The irq we extract from ACPI is not used - we deliver hypervisor interrupts on a special vector. Make the necessary adjustments. Signed-off-by:
-
K. Y. Srinivasan authored
The macro VMBUS_DEVICE() is unused; get rid of it. Signed-off-by:
-
K. Y. Srinivasan authored
Use uuid_le_cmp() for comparing GUIDs. Signed-off-by:
-
K. Y. Srinivasan authored
Consistently use uuid_le type in the Hyper-V driver code. Signed-off-by:
-
