platform/kernel/linux-rpi.git
20 months agovdpa/mlx5: Directly assign memory key
Eli Cohen [Sun, 5 Feb 2023 07:29:06 +0000 (09:29 +0200)]
vdpa/mlx5: Directly assign memory key

When creating a memory key, the key value should be assigned to the
passed pointer and not or'ed to.

No functional issue was observed due to this bug.

Fixes: 29064bfdabd5 ("vdpa/mlx5: Add support library for mlx5 VDPA implementation")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20230205072906.1108194-1-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agotools/virtio: enable to build with retpoline
Shunsuke Mie [Thu, 2 Feb 2023 10:45:38 +0000 (19:45 +0900)]
tools/virtio: enable to build with retpoline

Add build options to bring it close to a linux kernel. It allows for
testing that is close to reality.

Signed-off-by: Shunsuke Mie <mie@igel.co.jp>
Message-Id: <20230202104538.2041879-1-mie@igel.co.jp>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovringh: fix a typo in comments for vringh_kiov
Shunsuke Mie [Thu, 2 Feb 2023 10:42:48 +0000 (19:42 +0900)]
vringh: fix a typo in comments for vringh_kiov

Probably it is a simple copy error from struct vring_iov.

Signed-off-by: Shunsuke Mie <mie@igel.co.jp>
Message-Id: <20230202104248.2040652-1-mie@igel.co.jp>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovhost-vdpa: print warning when vhost_vdpa_alloc_domain fails
Alvaro Karsz [Thu, 2 Feb 2023 08:42:12 +0000 (10:42 +0200)]
vhost-vdpa: print warning when vhost_vdpa_alloc_domain fails

Add a print explaining why vhost_vdpa_alloc_domain failed if the device
is not IOMMU cache coherent capable.

Without this print, we have no hint why the operation failed.

For example:

$ virsh start <domain>
error: Failed to start domain <domain>
error: Unable to open '/dev/vhost-vdpa-<idx>' for vdpa device:
       Unknown error 524

Suggested-by: Eugenio Perez Martin <eperezma@redhat.com>
Signed-off-by: Alvaro Karsz <alvaro.karsz@solid-run.com>
Message-Id: <20230202084212.1328530-1-alvaro.karsz@solid-run.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Eugenio Pérez Martin <eperezma@redhat.com>
20 months agoscsi: virtio_scsi: fix handling of kmalloc failure
Zheng Wang [Thu, 2 Feb 2023 06:41:24 +0000 (14:41 +0800)]
scsi: virtio_scsi: fix handling of kmalloc failure

There is no check about the return value of kmalloc in
virtscsi_rescan_hotunplug. Add the check to avoid use
of null pointer 'inq_result' in case of the failure
of kmalloc.

Signed-off-by: Zheng Wang <zyytlz.wz@163.com>
Message-Id: <20230202064124.22277-1-zyytlz.wz@163.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa: Fix a couple of spelling mistakes in some messages
Colin Ian King [Mon, 30 Jan 2023 09:26:44 +0000 (09:26 +0000)]
vdpa: Fix a couple of spelling mistakes in some messages

There are two spelling mistakes in some literal strings. Fix them.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Message-Id: <20230130092644.37002-1-colin.i.king@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
20 months agovhost-net: support VIRTIO_F_RING_RESET
Kangjie Xu [Thu, 25 Aug 2022 08:56:10 +0000 (16:56 +0800)]
vhost-net: support VIRTIO_F_RING_RESET

Add VIRTIO_F_RING_RESET, which indicates that the driver can reset a
queue individually.

VIRTIO_F_RING_RESET feature is added to virtio-spec 1.2. The relevant
information is in
    oasis-tcs/virtio-spec#124
    oasis-tcs/virtio-spec#139

The implementation only adds the feature bit in supported features. It
does not require any other changes because we reuse the existing vhost
protocol.

The virtqueue reset process can be concluded as two parts:
1. The driver can reset a virtqueue. When it is triggered, we use the
set_backend to disable the virtqueue.
2. After the virtqueue is disabled, the driver may optionally re-enable
it. The process is basically similar to when the device is started,
except that the restart process does not need to set features and set
mem table since they do not change. QEMU will send messages containing
size, base, addr, kickfd and callfd of the virtqueue in order.
Specifically, the host kernel will receive these messages in order:
    a. VHOST_SET_VRING_NUM
    b. VHOST_SET_VRING_BASE
    c. VHOST_SET_VRING_ADDR
    d. VHOST_SET_VRING_KICK
    e. VHOST_SET_VRING_CALL
    f. VHOST_NET_SET_BACKEND
Finally, after we use set_backend to attach the virtqueue, the virtqueue
will be enabled and start to work.

Signed-off-by: Kangjie Xu <kangjie.xu@linux.alibaba.com>
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Message-Id: <20220825085610.80315-1-kangjie.xu@linux.alibaba.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
20 months agovhost-scsi: convert sysfs snprintf and sprintf to sysfs_emit
Bo Liu [Sun, 29 Jan 2023 09:11:45 +0000 (04:11 -0500)]
vhost-scsi: convert sysfs snprintf and sprintf to sysfs_emit

Follow the advice of the Documentation/filesystems/sysfs.rst
and show() should only use sysfs_emit() or sysfs_emit_at()
when formatting the value to be returned to user space.

Signed-off-by: Bo Liu <liubo03@inspur.com>
Message-Id: <20230129091145.2837-1-liubo03@inspur.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa: mlx5: support per virtqueue dma device
Jason Wang [Thu, 19 Jan 2023 06:15:25 +0000 (14:15 +0800)]
vdpa: mlx5: support per virtqueue dma device

This patch implements per virtqueue dma device for mlx5_vdpa. This is
needed for virtio_vdpa to work for CVQ which is backed by vringh but
not DMA. We simply advertise the vDPA device itself as the DMA device
for CVQ then DMA API can simply use PA so the identical mapping for
CVQ can still be used. Otherwise the identical (1:1) mapping won't
work when platform IOMMU is enabled since the IOVA is allocated on
demand which is not necessarily the PA.

This fixes the following crash when mlx5 vDPA device is bound to
virtio-vdpa with platform IOMMU enabled but not in passthrough mode:

BUG: unable to handle page fault for address: ff2fb3063deb1002
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 1393001067 P4D 1393002067 PUD 0
Oops: 0000 [#1] PREEMPT SMP NOPTI
CPU: 55 PID: 8923 Comm: kworker/u112:3 Kdump: loaded Not tainted 6.1.0+ #7
Hardware name: Dell Inc. PowerEdge R750/0PJ80M, BIOS 1.5.4 12/17/2021
Workqueue: mlx5_vdpa_wq mlx5_cvq_kick_handler [mlx5_vdpa]
RIP: 0010:vringh_getdesc_iotlb+0x93/0x1d0 [vringh]
Code: 14 25 40 ef 01 00 83 82 c0 0a 00 00 01 48 2b 05 93 5a 1b ea 8b 4c 24 14 48 c1 f8 06 48 c1 e0 0c 48 03 05 90 5a 1b ea 48 01 c8 <0f> b7 00 83 aa c0 0a 00 00 01 65 ff 0d bc e4 41 3f 0f 84 05 01 00
RSP: 0018:ff46821ba664fdf8 EFLAGS: 00010282
RAX: ff2fb3063deb1002 RBX: 0000000000000a20 RCX: 0000000000000002
RDX: ff2fb318d2f94380 RSI: 0000000000000002 RDI: 0000000000000001
RBP: ff2fb3065e832410 R08: ff46821ba664fe00 R09: 0000000000000001
R10: 0000000000000000 R11: 000000000000000d R12: ff2fb3065e832488
R13: ff2fb3065e8324a8 R14: ff2fb3065e8324c8 R15: ff2fb3065e8324a8
FS:  0000000000000000(0000) GS:ff2fb3257fac0000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ff2fb3063deb1002 CR3: 0000001392010006 CR4: 0000000000771ee0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
PKRU: 55555554
Call Trace:
<TASK>
  mlx5_cvq_kick_handler+0x89/0x2b0 [mlx5_vdpa]
  process_one_work+0x1e2/0x3b0
  ? rescuer_thread+0x390/0x390
  worker_thread+0x50/0x3a0
  ? rescuer_thread+0x390/0x390
  kthread+0xd6/0x100
  ? kthread_complete_and_exit+0x20/0x20
  ret_from_fork+0x1f/0x30
  </TASK>

Reviewed-by: Eli Cohen <elic@nvidia.com>
Tested-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230119061525.75068-6-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa: set dma mask for vDPA device
Jason Wang [Thu, 19 Jan 2023 06:15:24 +0000 (14:15 +0800)]
vdpa: set dma mask for vDPA device

Setting DMA mask for vDPA device in case that there are virtqueue that
is not backed by DMA so the vDPA device could be advertised as the DMA
device that is used by DMA API for software emulated virtqueues.

Reviewed-by: Eli Cohen <elic@nvidia.com>
Tested-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230119061525.75068-5-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovirtio-vdpa: support per vq dma device
Jason Wang [Thu, 19 Jan 2023 06:15:23 +0000 (14:15 +0800)]
virtio-vdpa: support per vq dma device

This patch adds the support of per vq dma device for virito-vDPA. vDPA
parents then are allowed to use different DMA devices. This is useful
for the parents that have software or emulated virtqueues.

Reviewed-by: Eli Cohen <elic@nvidia.com>
Tested-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230119061525.75068-4-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa: introduce get_vq_dma_device()
Jason Wang [Thu, 19 Jan 2023 06:15:22 +0000 (14:15 +0800)]
vdpa: introduce get_vq_dma_device()

This patch introduces a new method to query the dma device that is use
for a specific virtqueue.

Reviewed-by: Eli Cohen <elic@nvidia.com>
Tested-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230119061525.75068-3-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovirtio_ring: per virtqueue dma device
Jason Wang [Thu, 19 Jan 2023 06:15:21 +0000 (14:15 +0800)]
virtio_ring: per virtqueue dma device

This patch introduces a per virtqueue dma device. This will be used
for virtio devices whose virtqueue are backed by different underlayer
devices.

One example is the vDPA that where the control virtqueue could be
implemented through software mediation.

Some of the work are actually done before since the helper like
vring_dma_device(). This work left are:

- Let vring_dma_device() return the per virtqueue dma device instead
  of the vdev's parent.
- Allow passing a dma_device when creating the virtqueue through a new
  helper, old vring creation helper will keep using vdev's parent.

Reviewed-by: Eli Cohen <elic@nvidia.com>
Tested-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230119061525.75068-2-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovhost: remove unused paramete
Liming Wu [Tue, 10 Jan 2023 02:44:45 +0000 (10:44 +0800)]
vhost: remove unused paramete

"enabled" is defined in vhost_init_device_iotlb,
but it is never used. Let's remove it.

Signed-off-by: Liming Wu <liming.wu@jaguarmicro.com>
Message-Id: <20230110024445.303-1-liming.wu@jaguarmicro.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Eugenio Pérez <eperezma@redhat.com>
20 months agovhost-test: remove meaningless debug info
Liming Wu [Thu, 5 Jan 2023 07:03:56 +0000 (15:03 +0800)]
vhost-test: remove meaningless debug info

remove printk as it is meaningless.

Signed-off-by: Liming Wu <liming.wu@jaguarmicro.com>
Message-Id: <20230105070357.274-1-liming.wu@jaguarmicro.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
20 months agovdpa_sim: get rid of DMA ops
Jason Wang [Fri, 23 Dec 2022 06:00:21 +0000 (14:00 +0800)]
vdpa_sim: get rid of DMA ops

We used to (ab)use the DMA ops for setting up identical mappings in
the IOTLB. This patch tries to get rid of the those unnecessary DMA
ops by maintaining a simple identical/passthrough mappings by
default. When bound to virtio_vdpa driver, DMA API will simply use PA
as the IOVA and we will be all fine. When the vDPA bus tries to setup
customized mapping (e.g when bound to vhost-vDPA), the
identical/passthrough mapping will be removed.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20221223060021.28011-1-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Christoph Hellwig <hch@lst.de>
20 months agovdpa_sim_net: vendor satistics
Jason Wang [Fri, 23 Dec 2022 05:55:48 +0000 (13:55 +0800)]
vdpa_sim_net: vendor satistics

This patch adds support for basic vendor stats that include counters
for tx, rx and cvq.

Acked-by: Eugenio Pérez <eperezma@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20221223055548.27810-5-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
20 months agovdpa_sim: support vendor statistics
Jason Wang [Fri, 23 Dec 2022 05:55:47 +0000 (13:55 +0800)]
vdpa_sim: support vendor statistics

This patch adds a new config ops callback to allow individual
simulator to implement the vendor stats callback.

Acked-by: Eugenio Pérez <eperezma@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20221223055548.27810-4-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
20 months agovdpasim: customize allocation size
Jason Wang [Fri, 23 Dec 2022 05:55:46 +0000 (13:55 +0800)]
vdpasim: customize allocation size

Allow individual simulator to customize the allocation size.

Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Eugenio Pérez <eperezma@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20221223055548.27810-3-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa_sim: switch to use __vdpa_alloc_device()
Jason Wang [Fri, 23 Dec 2022 05:55:45 +0000 (13:55 +0800)]
vdpa_sim: switch to use __vdpa_alloc_device()

This allows us to control the allocation size of the structure.

Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Eugenio Pérez <eperezma@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20221223055548.27810-2-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa_sim: use weak barriers
Jason Wang [Wed, 21 Dec 2022 06:21:46 +0000 (14:21 +0800)]
vdpa_sim: use weak barriers

vDPA simulators are software emulated device, so let's switch to use
weak barriers to avoid extra overhead in the driver.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20221221062146.15356-1-jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Acked-by: Eugenio Pérez <eperezma@redhat.com>
20 months agovirtio-blk: support completion batching for the IRQ path
Suwan Kim [Wed, 21 Dec 2022 14:54:56 +0000 (23:54 +0900)]
virtio-blk: support completion batching for the IRQ path

This patch adds completion batching to the IRQ path. It reuses batch
completion code of virtblk_poll(). It collects requests to io_comp_batch
and processes them all at once. It can boost up the performance by 2%.

To validate the performance improvement and stabilty, I did fio test with
4 vCPU VM and 12 vCPU VM respectively. Both VMs have 8GB ram and the same
number of HW queues as vCPU.
The fio cammad is as follows and I ran the fio 5 times and got IOPS average.
(io_uring, randread, direct=1, bs=512, iodepth=64 numjobs=2,4)

Test result shows about 2% improvement.

           4 vcpu VM       |   numjobs=2   |   numjobs=4
      -----------------------------------------------------------
        fio without patch  |  367.2K IOPS  |   397.6K IOPS
      -----------------------------------------------------------
        fio with patch     |  372.8K IOPS  |   407.7K IOPS

           12 vcpu VM      |   numjobs=2   |   numjobs=4
      -----------------------------------------------------------
        fio without patch  |  363.6K IOPS  |   374.8K IOPS
      -----------------------------------------------------------
        fio with patch     |  373.8K IOPS  |   385.3K IOPS

Signed-off-by: Suwan Kim <suwan.kim027@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Message-Id: <20221221145456.281218-3-suwan.kim027@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovirtio-blk: set req->state to MQ_RQ_COMPLETE after polling I/O is finished
Suwan Kim [Wed, 21 Dec 2022 14:54:55 +0000 (23:54 +0900)]
virtio-blk: set req->state to MQ_RQ_COMPLETE after polling I/O is finished

Driver should set req->state to MQ_RQ_COMPLETE after it finishes to process
req. But virtio-blk doesn't set MQ_RQ_COMPLETE after virtblk_poll() handles
req and req->state still remains MQ_RQ_IN_FLIGHT. Fortunately so far there
is no issue about it because blk_mq_end_request_batch() sets req->state to
MQ_RQ_IDLE.

In this patch, virblk_poll() calls blk_mq_complete_request_remote() to set
req->state to MQ_RQ_COMPLETE before it adds req to a batch completion list.
So it properly sets req->state after polling I/O is finished.

Fixes: 4e0400525691 ("virtio-blk: support polling I/O")
Signed-off-by: Suwan Kim <suwan.kim027@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Message-Id: <20221221145456.281218-2-suwan.kim027@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agodocs: driver-api: virtio: commentize spec version checking
Bagas Sanjaya [Tue, 20 Dec 2022 09:58:29 +0000 (16:58 +0700)]
docs: driver-api: virtio: commentize spec version checking

A sentence that checks for later spec version is meant for developers
hacking the documentation source. Make it comment block (hidden from
actual output).

Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
Message-Id: <20221220095828.27588-4-bagasdotme@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agodocs: driver-api: virtio: slightly reword virtqueues allocation paragraph
Bagas Sanjaya [Tue, 20 Dec 2022 09:58:28 +0000 (16:58 +0700)]
docs: driver-api: virtio: slightly reword virtqueues allocation paragraph

"It's at this stage that" means "At this point", so use the latter as it
is more effective.

Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
Message-Id: <20221220095828.27588-3-bagasdotme@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agodocs: driver-api: virtio: parenthesize external reference targets
Bagas Sanjaya [Tue, 20 Dec 2022 09:58:27 +0000 (16:58 +0700)]
docs: driver-api: virtio: parenthesize external reference targets

Parenthesize targets to links in "References" section to distinguish
them from remaining texts.

While at it, describe the second target.

Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com>
Message-Id: <20221220095828.27588-2-bagasdotme@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa_sim: Implement resume vdpa op
Sebastien Boeuf [Tue, 3 Jan 2023 10:51:08 +0000 (11:51 +0100)]
vdpa_sim: Implement resume vdpa op

Implement resume operation for vdpa_sim devices, so vhost-vdpa will
offer that backend feature and userspace can effectively resume the
device.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Message-Id: <15a4566826033c5dd9a2167e5cfb0ef4d90cea49.1672742878.git.sebastien.boeuf@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
20 months agovhost-vdpa: uAPI to resume the device
Sebastien Boeuf [Tue, 3 Jan 2023 10:51:07 +0000 (11:51 +0100)]
vhost-vdpa: uAPI to resume the device

This new ioctl adds support for resuming the device from userspace.

This is required when trying to restore the device in a functioning
state after it's been suspended. It is already possible to reset a
suspended device, but that means the device must be reconfigured and
all the IOMMU/IOTLB mappings must be recreated. This new operation
allows the device to be resumed without going through a full reset.

This is particularly useful when trying to perform offline migration of
a virtual machine (also known as snapshot/restore) as it allows the VMM
to resume the virtual machine back to a running state after the snapshot
is performed.

Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Message-Id: <73b75fb87d25cff59768b4955a81fe7ffe5b4770.1672742878.git.sebastien.boeuf@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
20 months agovhost-vdpa: Introduce RESUME backend feature bit
Sebastien Boeuf [Tue, 3 Jan 2023 10:51:06 +0000 (11:51 +0100)]
vhost-vdpa: Introduce RESUME backend feature bit

Userspace knows if the device can be resumed or not by checking this
feature bit.

It's only exposed if the vdpa driver backend implements the resume()
operation callback. Userspace trying to negotiate this feature when it
hasn't been exposed will result in an error.

Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Message-Id: <b18db236ba3d990cdb41278eb4703be9201d9514.1672742878.git.sebastien.boeuf@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
20 months agovdpa: Add resume operation
Sebastien Boeuf [Tue, 3 Jan 2023 10:51:05 +0000 (11:51 +0100)]
vdpa: Add resume operation

Add a new operation to allow a vDPA device to be resumed after it has
been suspended. Trying to resume a device that wasn't suspended will
result in a no-op.

This operation is optional. If it's not implemented, the associated
backend feature bit will not be exposed. And if the feature bit is not
exposed, invoking this operation will return an error.

Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Message-Id: <6e05c4b31b47f3e29cb2bd7ebd56c81f84b8f48a.1672742878.git.sebastien.boeuf@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
20 months agovirtio: vdpa: new SolidNET DPU driver.
Alvaro Karsz [Tue, 10 Jan 2023 16:56:38 +0000 (18:56 +0200)]
virtio: vdpa: new SolidNET DPU driver.

This commit includes:
 1) The driver to manage the controlplane over vDPA bus.
 2) A HW monitor device to read health values from the DPU.

Signed-off-by: Alvaro Karsz <alvaro.karsz@solid-run.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230110165638.123745-4-alvaro.karsz@solid-run.com>
Message-Id: <20230209075128.78915-1-alvaro.karsz@solid-run.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agoPCI: Avoid FLR for SolidRun SNET DPU rev 1
Alvaro Karsz [Tue, 10 Jan 2023 16:56:37 +0000 (18:56 +0200)]
PCI: Avoid FLR for SolidRun SNET DPU rev 1

This patch fixes a FLR bug on the SNET DPU rev 1 by setting the
PCI_DEV_FLAGS_NO_FLR_RESET flag.

As there is a quirk to avoid FLR (quirk_no_flr), I added a new quirk
to check the rev ID before calling to quirk_no_flr.

Without this patch, a SNET DPU rev 1 may hang when FLR is applied.

Signed-off-by: Alvaro Karsz <alvaro.karsz@solid-run.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Message-Id: <20230110165638.123745-3-alvaro.karsz@solid-run.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agoPCI: Add SolidRun vendor ID
Alvaro Karsz [Tue, 10 Jan 2023 16:56:36 +0000 (18:56 +0200)]
PCI: Add SolidRun vendor ID

Add SolidRun vendor ID to pci_ids.h

The vendor ID is used in 2 different source files, the SNET vDPA driver
and PCI quirks.

Signed-off-by: Alvaro Karsz <alvaro.karsz@solid-run.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Message-Id: <20230110165638.123745-2-alvaro.karsz@solid-run.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa_sim_net: Offer VIRTIO_NET_F_STATUS
Eugenio Pérez [Thu, 17 Nov 2022 15:55:02 +0000 (16:55 +0100)]
vdpa_sim_net: Offer VIRTIO_NET_F_STATUS

VIRTIO_NET_S_LINK_UP is already returned in config reads since vdpasim
creation, but the feature bit was not offered to the driver.

Tested modifying VIRTIO_NET_S_LINK_UP and different values of "status"
in qemu virtio-net options, using vhost_vdpa.

Not considering as a fix, because there should be no driver trusting in
this config read before the feature flag.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20221117155502.1394700-1-eperezma@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
20 months agovDPA/ifcvf: implement features provisioning
Zhu Lingshan [Fri, 25 Nov 2022 14:57:24 +0000 (22:57 +0800)]
vDPA/ifcvf: implement features provisioning

This commit implements features provisioning for ifcvf, that means:
1)checkk whether the provisioned features are supported by
the management device
2)vDPA device only presents selected feature bits

Examples:
a)The management device supported features:
$ vdpa mgmtdev show pci/0000:01:00.5
pci/0000:01:00.5:
  supported_classes net
  max_supported_vqs 9
  dev_features MTU MAC MRG_RXBUF CTRL_VQ MQ ANY_LAYOUT VERSION_1 ACCESS_PLATFORM

b)Provision a vDPA device with all supported features:
$ vdpa dev add name vdpa0 mgmtdev pci/0000:01:00.5
$ vdpa/vdpa dev config show vdpa0
vdpa0: mac 00:e8:ca:11:be:05 link up link_announce false max_vq_pairs 4 mtu 1500
  negotiated_features MRG_RXBUF CTRL_VQ MQ VERSION_1 ACCESS_PLATFORM

c)Provision a vDPA device with a subset of the supported features:
$ vdpa dev add name vdpa0 mgmtdev pci/0000:01:00.5 device_features 0x300020020
$ vdpa dev config show vdpa0
mac 00:e8:ca:11:be:05 link up link_announce false
  negotiated_features CTRL_VQ VERSION_1 ACCESS_PLATFORM

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Message-Id: <20221125145724.1129962-13-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: retire ifcvf_private_to_vf
Zhu Lingshan [Fri, 25 Nov 2022 14:57:23 +0000 (22:57 +0800)]
vDPA/ifcvf: retire ifcvf_private_to_vf

This commit retires ifcvf_private_to_vf, because
the vf is already a member of the adapter,
so it could be easily addressed by adapter->vf.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Message-Id: <20221125145724.1129962-12-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: allocate the adapter in dev_add()
Zhu Lingshan [Fri, 25 Nov 2022 14:57:22 +0000 (22:57 +0800)]
vDPA/ifcvf: allocate the adapter in dev_add()

The adapter is the container of the vdpa_device,
this commits allocate the adapter in dev_add()
rather than in probe(). So that the vdpa_device()
could be re-created when the userspace creates
the vdpa device, and free-ed in dev_del()

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-11-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: manage ifcvf_hw in the mgmt_dev
Zhu Lingshan [Fri, 25 Nov 2022 14:57:21 +0000 (22:57 +0800)]
vDPA/ifcvf: manage ifcvf_hw in the mgmt_dev

This commit allocates the hw structure in the
management device structure. So the hardware
can be initialized once the management device
is allocated in probe.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-10-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: ifcvf_request_irq works on ifcvf_hw
Zhu Lingshan [Fri, 25 Nov 2022 14:57:20 +0000 (22:57 +0800)]
vDPA/ifcvf: ifcvf_request_irq works on ifcvf_hw

All ifcvf_request_irq's callees are refactored
to work on ifcvf_hw, so it should be decoupled
from the adapter as well

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-9-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: decouple config/dev IRQ requester and vectors allocator from the adapter
Zhu Lingshan [Fri, 25 Nov 2022 14:57:19 +0000 (22:57 +0800)]
vDPA/ifcvf: decouple config/dev IRQ requester and vectors allocator from the adapter

This commit decouples the config irq requester, the device
shared irq requester and the MSI vectors allocator from
the adapter. So they can be safely invoked since probe
before the adapter is allocated.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-8-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: decouple vq irq requester from the adapter
Zhu Lingshan [Fri, 25 Nov 2022 14:57:18 +0000 (22:57 +0800)]
vDPA/ifcvf: decouple vq irq requester from the adapter

This commit decouples the vq irq requester from the adapter,
so that these functions can be invoked since probe.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-7-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: decouple config IRQ releaser from the adapter
Zhu Lingshan [Fri, 25 Nov 2022 14:57:17 +0000 (22:57 +0800)]
vDPA/ifcvf: decouple config IRQ releaser from the adapter

This commit decouples config IRQ releaser from the adapter,
so that it could be invoked once probe or in err handlers.
ifcvf_free_irq() works on ifcvf_hw in this commit

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-6-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: decouple vq IRQ releasers from the adapter
Zhu Lingshan [Fri, 25 Nov 2022 14:57:16 +0000 (22:57 +0800)]
vDPA/ifcvf: decouple vq IRQ releasers from the adapter

This commit decouples the IRQ releasers from the
adapter, so that these functions could be
safely invoked once probe

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-5-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: alloc the mgmt_dev before the adapter
Zhu Lingshan [Fri, 25 Nov 2022 14:57:15 +0000 (22:57 +0800)]
vDPA/ifcvf: alloc the mgmt_dev before the adapter

This commit reverses the order of allocating the
management device and the adapter. So that it would
be possible to move the allocation of the adapter
to dev_add().

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-4-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: decouple config space ops from the adapter
Zhu Lingshan [Fri, 25 Nov 2022 14:57:14 +0000 (22:57 +0800)]
vDPA/ifcvf: decouple config space ops from the adapter

This commit decopules the config space ops from the
adapter layer, so these functions can be invoked
once the device is probed.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-3-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovDPA/ifcvf: decouple hw features manipulators from the adapter
Zhu Lingshan [Fri, 25 Nov 2022 14:57:13 +0000 (22:57 +0800)]
vDPA/ifcvf: decouple hw features manipulators from the adapter

This commit gets rid of ifcvf_adapter in hw features related
functions in ifcvf_base. Then these functions are more rubust
and de-coupling from the ifcvf_adapter layer. So these
functions could be invoded once the device is probed, even
before the adapter is allocaed.

Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
Cc: stable@vger.kernel.org
Message-Id: <20221125145724.1129962-2-lingshan.zhu@intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa/mlx5: Add RX counters to debugfs
Eli Cohen [Mon, 14 Nov 2022 13:17:59 +0000 (15:17 +0200)]
vdpa/mlx5: Add RX counters to debugfs

For each interface, either VLAN tagged or untagged, add two hardware
counters: one for unicast and another for multicast. The counters count
RX packets and bytes and can be read through debugfs:

$ cat /sys/kernel/debug/mlx5/mlx5_core.sf.1/vdpa-0/rx/untagged/mcast/packets
$ cat /sys/kernel/debug/mlx5/mlx5_core.sf.1/vdpa-0/rx/untagged/ucast/bytes

This feature is controlled via the config option
MLX5_VDPA_STEERING_DEBUG. It is off by default as it may have some
impact on performance.

includes a fixup By Yang Yingliang <yangyingliang@huawei.com>:

vdpa/mlx5: fix check wrong pointer in mlx5_vdpa_add_mac_vlan_rules()

The local variable 'rule' is not used anymore, fix return value
check after calling mlx5_add_flow_rules().

Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-9-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Message-Id: <20230104074418.1737510-1-yangyingliang@huawei.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Eli Cohen <elic@nvidia.com>
Acked-by: Jason Wang <jasowang@redhat.com>
20 months agovdpa/mlx5: Add debugfs subtree
Eli Cohen [Mon, 14 Nov 2022 13:17:58 +0000 (15:17 +0200)]
vdpa/mlx5: Add debugfs subtree

Add debugfs subtree and expose flow table ID and TIR number. This
information can be used by external tools to do extended
troubleshooting.

The information can be retrieved like so:
$ cat /sys/kernel/debug/mlx5/mlx5_core.sf.1/vdpa-0/rx/table_id
$ cat /sys/kernel/debug/mlx5/mlx5_core.sf.1/vdpa-0/rx/tirn

Reviewed-by: Si-Wei Liu <si-wei.liu@oracle.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-8-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa/mlx5: Move some definitions to a new header file
Eli Cohen [Mon, 14 Nov 2022 13:17:57 +0000 (15:17 +0200)]
vdpa/mlx5: Move some definitions to a new header file

Move some definitions from mlx5_vnet.c to newly added header file
mlx5_vnet.h. We need these definitions for the following patches that
add debugfs tree to expose information vital for debug.

Reviewed-by: Si-Wei Liu <si-wei.liu@oracle.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Message-Id: <20221114131759.57883-7-elic@nvidia.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovirtio_blk: mark all zone fields LE
Michael S. Tsirkin [Thu, 22 Dec 2022 19:30:13 +0000 (14:30 -0500)]
virtio_blk: mark all zone fields LE

zone is a virtio 1.x feature so all fields are LE,
they are handled as such, but have mistakenly been labeled
__virtioXX in the header.  This results in a bunch of sparse warnings.

Use the __leXX tags to make sparse happy.

Message-Id: <20221222193214.55146-1-mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovirtio_blk: zone append in header type tweak
Michael S. Tsirkin [Tue, 20 Dec 2022 12:37:49 +0000 (07:37 -0500)]
virtio_blk: zone append in header type tweak

virtio blk returns a 64 bit append_sector in an input buffer,
in LE format. This field is not tagged as LE correctly, so
even though the generated code is ok, we get warnings from sparse:

drivers/block/virtio_blk.c:332:33: sparse: sparse: cast to restricted __le64

Make sparse happy by using the correct type.

Message-Id: <20221220125154.564265-1-mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovirtio_blk: temporary variable type tweak
Michael S. Tsirkin [Tue, 20 Dec 2022 12:37:49 +0000 (07:37 -0500)]
virtio_blk: temporary variable type tweak

virtblk_result returns blk_status_t which is a bitwise restricted type,
so we are not supposed to stuff it in a plain int temporary variable.
All we do with it is pass it on to a function expecting blk_status_t so
the generated code is ok, but we get warnings from sparse:

drivers/block/virtio_blk.c:326:36: sparse: sparse: incorrect type in initializer (different base types) @@     expected int status @@
+got restricted blk_status_t @@
drivers/block/virtio_blk.c:334:33: sparse: sparse: incorrect type in argument 2 (different base types) @@     expected restricted
+blk_status_t [usertype] error @@     got int status @@

Make sparse happy by using the correct type.

Message-Id: <20221220124152.523531-1-mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
20 months agovirtio-blk: add support for zoned block devices
Dmitry Fomichev [Sun, 16 Oct 2022 03:41:27 +0000 (23:41 -0400)]
virtio-blk: add support for zoned block devices

This patch adds support for Zoned Block Devices (ZBDs) to the kernel
virtio-blk driver.

The patch accompanies the virtio-blk ZBD support draft that is now
being proposed for standardization. The latest version of the draft is
linked at

https://github.com/oasis-tcs/virtio-spec/issues/143 .

The QEMU zoned device code that implements these protocol extensions
has been developed by Sam Li and it is currently in review at the QEMU
mailing list.

A number of virtblk request structure changes has been introduced to
accommodate the functionality that is specific to zoned block devices
and, most importantly, make room for carrying the Zoned Append sector
value from the device back to the driver along with the request status.

The zone-specific code in the patch is heavily influenced by NVMe ZNS
code in drivers/nvme/host/zns.c, but it is simpler because the proposed
virtio ZBD draft only covers the zoned device features that are
relevant to the zoned functionality provided by Linux block layer.

includes the following fixup:

virtio-blk: fix probe without CONFIG_BLK_DEV_ZONED

When building without CONFIG_BLK_DEV_ZONED, VIRTIO_BLK_F_ZONED
is excluded from array of driver features.
As a result virtio_has_feature panics in virtio_check_driver_offered_feature
since that by design verifies that a feature we are checking for
is listed in the feature array.

To fix, replace the call to virtio_has_feature with a stub.

Message-Id: <20221016034127.330942-3-dmitry.fomichev@wdc.com>
Co-developed-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Dmitry Fomichev <dmitry.fomichev@wdc.com>
Message-Id: <20221220112340.518841-1-mst@redhat.com>
Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Reported-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Debugged-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Tested-by: Anders Roxell <anders.roxell@linaro.org>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
20 months agodocs: driver-api: virtio: virtio on Linux
Ricardo Cañuelo [Mon, 10 Oct 2022 06:43:59 +0000 (08:43 +0200)]
docs: driver-api: virtio: virtio on Linux

Basic doc about Virtio on Linux and a short tutorial on Virtio drivers.

includes the following fixup:

virtio: fix virtio_config_ops kerneldocs

Fixes two warning messages when building htmldocs:

    warning: duplicate section name 'Note'
    warning: expecting prototype for virtio_config_ops().
             Prototype was for vq_callback_t() instead

Message-Id: <20221010064359.1324353-2-ricardo.canuelo@collabora.com>
Signed-off-by: Ricardo Cañuelo <ricardo.canuelo@collabora.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <20221220100035.2712449-1-ricardo.canuelo@collabora.com>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Reviewed-by: Bagas Sanjaya <bagasdotme@gmail.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovirtio_pmem: populate numa information
Michael Sammler [Tue, 15 Nov 2022 21:40:36 +0000 (21:40 +0000)]
virtio_pmem: populate numa information

Compute the numa information for a virtio_pmem device from the memory
range of the device. Previously, the target_node was always 0 since
the ndr_desc.target_node field was never explicitly set. The code for
computing the numa node is taken from cxl_pmem_region_probe in
drivers/cxl/pmem.c.

Signed-off-by: Michael Sammler <sammler@google.com>
Reviewed-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Tested-by: Mina Almasry <almasrymina@google.com>
Message-Id: <20221115214036.1571015-1-sammler@google.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
20 months agovdpa_sim: not reset state in vdpasim_queue_ready
Eugenio Pérez [Wed, 18 Jan 2023 16:43:58 +0000 (17:43 +0100)]
vdpa_sim: not reset state in vdpasim_queue_ready

vdpasim_queue_ready calls vringh_init_iotlb, which resets split indexes.
But it can be called after setting a ring base with
vdpasim_set_vq_state.

Fix it by stashing them. They're still resetted in vdpasim_vq_reset.

This was discovered and tested live migrating the vdpa_sim_net device.

Fixes: 2c53d0f64c06 ("vdpasim: vDPA device simulator")
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230118164359.1523760-2-eperezma@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
20 months agoLinux 6.2-rc8
Linus Torvalds [Sun, 12 Feb 2023 22:10:17 +0000 (14:10 -0800)]
Linux 6.2-rc8

20 months agoMAINTAINERS: Add myself as maintainer for arch/sh (SUPERH)
John Paul Adrian Glaubitz [Tue, 7 Feb 2023 16:57:15 +0000 (17:57 +0100)]
MAINTAINERS: Add myself as maintainer for arch/sh (SUPERH)

Both Rich Felker and Yoshinori Sato haven't done any work on arch/sh
for a while. As I have been maintaining Debian's sh4 port since 2014,
I am interested to keep the architecture alive.

Signed-off-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Acked-by: Yoshinori Sato <ysato@users.sourceforge.jp>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
20 months agoMerge tag 'trace-v6.2-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/trace...
Linus Torvalds [Sun, 12 Feb 2023 21:52:17 +0000 (13:52 -0800)]
Merge tag 'trace-v6.2-rc7' of git://git./linux/kernel/git/trace/linux-trace

Pull tracing fix from Steven Rostedt:
 "Fix showing of TASK_COMM_LEN instead of its value

  The TASK_COMM_LEN was converted from a macro into an enum so that BTF
  would have access to it. But this unfortunately caused TASK_COMM_LEN
  to display in the format fields of trace events, as they are created
  by the TRACE_EVENT() macro and such, macros convert to their values,
  where as enums do not.

  To handle this, instead of using the field itself to be display, save
  the value of the array size as another field in the trace_event_fields
  structure, and use that instead.

  Not only does this fix the issue, but also converts the other trace
  events that have this same problem (but were not breaking tooling).

  With this change, the original work around b3bc8547d3be6 ("tracing:
  Have TRACE_DEFINE_ENUM affect trace event types as well") could be
  reverted (but that should be done in the merge window)"

* tag 'trace-v6.2-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
  tracing: Fix TASK_COMM_LEN in trace event format file

20 months agoMerge tag 'for-6.2-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Sun, 12 Feb 2023 19:26:36 +0000 (11:26 -0800)]
Merge tag 'for-6.2-rc7-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - one more fix for a tree-log 'write time corruption' report, update
   the last dir index directly and don't keep in the log context

 - do VFS-level inode lock around FIEMAP to prevent a deadlock with
   concurrent fsync, the extent-level lock is not sufficient

 - don't cache a single-device filesystem device to avoid cases when a
   loop device is reformatted and the entry gets stale

* tag 'for-6.2-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: free device in btrfs_close_devices for a single device filesystem
  btrfs: lock the inode in shared mode before starting fiemap
  btrfs: simplify update of last_dir_index_offset when logging a directory

20 months agoMerge tag 'usb-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sun, 12 Feb 2023 19:18:57 +0000 (11:18 -0800)]
Merge tag 'usb-6.2-rc8' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are 2 small USB driver fixes that resolve some reported
  regressions and one new device quirk. Specifically these are:

   - new quirk for Alcor Link AK9563 smartcard reader

   - revert of u_ether gadget change in 6.2-rc1 that caused problems

   - typec pin probe fix

  All of these have been in linux-next with no reported problems"

* tag 'usb-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: core: add quirk for Alcor Link AK9563 smartcard reader
  usb: typec: altmodes/displayport: Fix probe pin assign check
  Revert "usb: gadget: u_ether: Do not make UDC parent of the net device"

20 months agoMerge tag 'efi-fixes-for-v6.2-4' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 12 Feb 2023 19:13:29 +0000 (11:13 -0800)]
Merge tag 'efi-fixes-for-v6.2-4' of git://git./linux/kernel/git/efi/efi

Pull EFI fix from Ard Biesheuvel:
 "A fix from Darren to widen the SMBIOS match for detecting Ampere Altra
  machines with problematic firmware. In the mean time, we are working
  on a more precise check, but this is still work in progress"

* tag 'efi-fixes-for-v6.2-4' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi:
  arm64: efi: Force the use of SetVirtualAddressMap() on eMAG and Altra Max machines

20 months agoMerge tag 'powerpc-6.2-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sun, 12 Feb 2023 19:08:15 +0000 (11:08 -0800)]
Merge tag 'powerpc-6.2-5' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - Fix interrupt exit race with security mitigation switching.

 - Don't select ARCH_WANTS_NO_INSTR until warnings are fixed.

 - Build fix for CONFIG_NUMA=n.

Thanks to Nicholas Piggin, Randy Dunlap, and Sachin Sant.

* tag 'powerpc-6.2-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/64s/interrupt: Fix interrupt exit race with security mitigation switch
  powerpc/kexec_file: fix implicit decl error
  powerpc: Don't select ARCH_WANTS_NO_INSTR

20 months agoFix page corruption caused by racy check in __free_pages
David Chen [Thu, 9 Feb 2023 17:48:28 +0000 (17:48 +0000)]
Fix page corruption caused by racy check in __free_pages

When we upgraded our kernel, we started seeing some page corruption like
the following consistently:

  BUG: Bad page state in process ganesha.nfsd  pfn:1304ca
  page:0000000022261c55 refcount:0 mapcount:-128 mapping:0000000000000000 index:0x0 pfn:0x1304ca
  flags: 0x17ffffc0000000()
  raw: 0017ffffc0000000 ffff8a513ffd4c98 ffffeee24b35ec08 0000000000000000
  raw: 0000000000000000 0000000000000001 00000000ffffff7f 0000000000000000
  page dumped because: nonzero mapcount
  CPU: 0 PID: 15567 Comm: ganesha.nfsd Kdump: loaded Tainted: P    B      O      5.10.158-1.nutanix.20221209.el7.x86_64 #1
  Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 04/05/2016
  Call Trace:
   dump_stack+0x74/0x96
   bad_page.cold+0x63/0x94
   check_new_page_bad+0x6d/0x80
   rmqueue+0x46e/0x970
   get_page_from_freelist+0xcb/0x3f0
   ? _cond_resched+0x19/0x40
   __alloc_pages_nodemask+0x164/0x300
   alloc_pages_current+0x87/0xf0
   skb_page_frag_refill+0x84/0x110
   ...

Sometimes, it would also show up as corruption in the free list pointer
and cause crashes.

After bisecting the issue, we found the issue started from commit
e320d3012d25 ("mm/page_alloc.c: fix freeing non-compound pages"):

if (put_page_testzero(page))
free_the_page(page, order);
else if (!PageHead(page))
while (order-- > 0)
free_the_page(page + (1 << order), order);

So the problem is the check PageHead is racy because at this point we
already dropped our reference to the page.  So even if we came in with
compound page, the page can already be freed and PageHead can return
false and we will end up freeing all the tail pages causing double free.

Fixes: e320d3012d25 ("mm/page_alloc.c: fix freeing non-compound pages")
Link: https://lore.kernel.org/lkml/BYAPR02MB448855960A9656EEA81141FC94D99@BYAPR02MB4488.namprd02.prod.outlook.com/
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@vger.kernel.org
Signed-off-by: Chunwei Chen <david.chen@nutanix.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
20 months agotracing: Fix TASK_COMM_LEN in trace event format file
Yafang Shao [Sun, 12 Feb 2023 15:13:03 +0000 (15:13 +0000)]
tracing: Fix TASK_COMM_LEN in trace event format file

After commit 3087c61ed2c4 ("tools/testing/selftests/bpf: replace open-coded 16 with TASK_COMM_LEN"),
the content of the format file under
/sys/kernel/tracing/events/task/task_newtask was changed from
  field:char comm[16];    offset:12;    size:16;    signed:0;
to
  field:char comm[TASK_COMM_LEN];    offset:12;    size:16;    signed:0;

John reported that this change breaks older versions of perfetto.
Then Mathieu pointed out that this behavioral change was caused by the
use of __stringify(_len), which happens to work on macros, but not on enum
labels. And he also gave the suggestion on how to fix it:
  :One possible solution to make this more robust would be to extend
  :struct trace_event_fields with one more field that indicates the length
  :of an array as an actual integer, without storing it in its stringified
  :form in the type, and do the formatting in f_show where it belongs.

The result as follows after this change,
$ cat /sys/kernel/tracing/events/task/task_newtask/format
        field:char comm[16];    offset:12;      size:16;        signed:0;

Link: https://lore.kernel.org/lkml/Y+QaZtz55LIirsUO@google.com/
Link: https://lore.kernel.org/linux-trace-kernel/20230210155921.4610-1-laoar.shao@gmail.com/
Link: https://lore.kernel.org/linux-trace-kernel/20230212151303.12353-1-laoar.shao@gmail.com
Cc: stable@vger.kernel.org
Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Kajetan Puchalski <kajetan.puchalski@arm.com>
CC: Qais Yousef <qyousef@layalina.io>
Fixes: 3087c61ed2c4 ("tools/testing/selftests/bpf: replace open-coded 16 with TASK_COMM_LEN")
Reported-by: John Stultz <jstultz@google.com>
Debugged-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Suggested-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
21 months agoMerge tag 'spi-fix-v6.2-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi
Linus Torvalds [Sun, 12 Feb 2023 00:39:05 +0000 (16:39 -0800)]
Merge tag 'spi-fix-v6.2-rc7' of git://git./linux/kernel/git/broonie/spi

Pull spi fixes from Mark Brown:
 "A couple of hopefully final fixes for spi: one driver specific fix for
  an issue with very large transfers and a fix for an issue with the
  locking fixes in spidev merged earlier this release cycle which was
  missed"

* tag 'spi-fix-v6.2-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
  spi: spidev: fix a recursive locking error
  spi: dw: Fix wrong FIFO level setting for long xfers

21 months agoMerge tag 'x86-urgent-2023-02-11' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 11 Feb 2023 19:17:21 +0000 (11:17 -0800)]
Merge tag 'x86-urgent-2023-02-11' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "Fix a kprobes bug, plus add a new Intel model number to the upstream
  <asm/intel-family.h> header for drivers to use"

* tag 'x86-urgent-2023-02-11' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/cpu: Add Lunar Lake M
  x86/kprobes: Fix 1 byte conditional jump target

21 months agoMerge tag 'locking-urgent-2023-02-11' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 11 Feb 2023 19:11:18 +0000 (11:11 -0800)]
Merge tag 'locking-urgent-2023-02-11' of git://git./linux/kernel/git/tip/tip

Pull locking fix from Ingo Molnar:
 "Fix an rtmutex missed-wakeup bug"

* tag 'locking-urgent-2023-02-11' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  rtmutex: Ensure that the top waiter is always woken up

21 months agoMerge tag 'cxl-fixes-6.2' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl
Linus Torvalds [Sat, 11 Feb 2023 19:03:25 +0000 (11:03 -0800)]
Merge tag 'cxl-fixes-6.2' of git://git./linux/kernel/git/cxl/cxl

Pull cxl fixes from Dan Williams:
 "Two fixups for CXL (Compute Express Link) in presence of passthrough
  decoders.

  This primarily helps developers using the QEMU CXL emulation, but with
  the impending arrival of CXL switches these types of topologies will
  be of interest to end users.

   - Fix a crash when shutting down regions in the presence of
     passthrough decoders

   - Fix region creation to understand passthrough decoders instead of
     the narrower definition of passthrough ports"

* tag 'cxl-fixes-6.2' of git://git.kernel.org/pub/scm/linux/kernel/git/cxl/cxl:
  cxl/region: Fix passthrough-decoder detection
  cxl/region: Fix null pointer dereference for resetting decoder

21 months agoMerge tag 'libnvdimm-fixes-6.2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 11 Feb 2023 18:58:36 +0000 (10:58 -0800)]
Merge tag 'libnvdimm-fixes-6.2' of git://git./linux/kernel/git/nvdimm/nvdimm

Pull libnvdimm fixes from Dan Williams:
 "A fix for an issue that could causes users to inadvertantly reserve
  too much capacity when debugging the KMSAN and persistent memory
  namespace, a lockdep fix, and a kernel-doc build warning:

   - Resolve the conflict between KMSAN and NVDIMM with respect to
     reserving pmem namespace / volume capacity for larger sizeof(struct
     page)

   - Fix a lockdep warning in the the NFIT code

   - Fix a kernel-doc build warning"

* tag 'libnvdimm-fixes-6.2' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  nvdimm: Support sizeof(struct page) > MAX_STRUCT_PAGE_SIZE
  ACPI: NFIT: fix a potential deadlock during NFIT teardown
  dax: super.c: fix kernel-doc bad line warning

21 months agoMerge tag 'fixes-2023-02-11' of git://git.kernel.org/pub/scm/linux/kernel/git/rppt...
Linus Torvalds [Sat, 11 Feb 2023 18:51:45 +0000 (10:51 -0800)]
Merge tag 'fixes-2023-02-11' of git://git./linux/kernel/git/rppt/memblock

Pull memblock revert from Mike Rapoport:
 "Revert 'mm: Always release pages to the buddy allocator in
  memblock_free_late()'

  The pages being freed by memblock_free_late() have already been
  initialized, but if they are in the deferred init range,
  __free_one_page() might access nearby uninitialized pages when trying
  to coalesce buddies, which will cause a crash.

  A proper fix will be more involved so revert this change for the time
  being"

* tag 'fixes-2023-02-11' of git://git.kernel.org/pub/scm/linux/kernel/git/rppt/memblock:
  Revert "mm: Always release pages to the buddy allocator in memblock_free_late()."

21 months agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 10 Feb 2023 23:28:08 +0000 (15:28 -0800)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fixes from Stephen Boyd:
 "Two clk driver fixes

   - Use devm_kasprintf() to avoid overflows when forming clk names in
     the Microchip PolarFire driver

   - Fix the pretty broken Ingenic JZ4760 M/N/OD calculation to actually
     work and find proper divisors"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: ingenic: jz4760: Update M/N/OD calculation algorithm
  clk: microchip: mpfs-ccc: Use devm_kasprintf() for allocating formatted strings

21 months agoMerge tag 'pinctrl-v6.2-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Fri, 10 Feb 2023 23:02:16 +0000 (15:02 -0800)]
Merge tag 'pinctrl-v6.2-3' of git://git./linux/kernel/git/linusw/linux-pinctrl

Pull pin control fixes from Linus Walleij:
 "Some assorted pin control fixes, the most interesting will be the
  Intel patch fixing a classic problem: laptop touchpad IRQs...

   - Some pin drive register fixes in the Mediatek driver.

   - Return proper error code in the Aspeed driver, and revert and
     ill-advised force-disablement patch that needs to be reworked.

   - Fix AMD driver debug output.

   - Fix potential NULL dereference in the Single driver.

   - Fix a group definition error in the Qualcomm SM8450 LPASS driver.

   - Restore pins used in direct IRQ mode in the Intel driver (This
     fixes some laptop touchpads!)"

* tag 'pinctrl-v6.2-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: intel: Restore the pins that used to be in Direct IRQ mode
  pinctrl: qcom: sm8450-lpass-lpi: correct swr_rx_data group
  pinctrl: aspeed: Revert "Force to disable the function's signal"
  pinctrl: single: fix potential NULL dereference
  pinctrl: amd: Fix debug output for debounce time
  pinctrl: aspeed: Fix confusing types in return value
  pinctrl: mediatek: Fix the drive register definition of some Pins

21 months agoMerge tag 'pci-v6.2-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci
Linus Torvalds [Fri, 10 Feb 2023 22:18:48 +0000 (14:18 -0800)]
Merge tag 'pci-v6.2-fixes-2' of git://git./linux/kernel/git/pci/pci

Pull PCI fixes from Bjorn Helgaas:

 - Move to a shared PCI git tree (Bjorn Helgaas)

 - Add Krzysztof Wilczyński as another PCI maintainer (Lorenzo
   Pieralisi)

 - Revert a couple ASPM patches to fix suspend/resume regressions (Bjorn
   Helgaas)

* tag 'pci-v6.2-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci:
  Revert "PCI/ASPM: Refactor L1 PM Substates Control Register programming"
  Revert "PCI/ASPM: Save L1 PM Substates Capability for suspend/resume"
  MAINTAINERS: Promote Krzysztof to PCI controller maintainer
  MAINTAINERS: Move to shared PCI tree

21 months agoRevert "PCI/ASPM: Refactor L1 PM Substates Control Register programming"
Bjorn Helgaas [Fri, 3 Feb 2023 19:57:39 +0000 (13:57 -0600)]
Revert "PCI/ASPM: Refactor L1 PM Substates Control Register programming"

This reverts commit 5e85eba6f50dc288c22083a7e213152bcc4b8208.

Thomas Witt reported that 5e85eba6f50d ("PCI/ASPM: Refactor L1 PM Substates
Control Register programming") broke suspend/resume on a Tuxedo
Infinitybook S 14 v5, which seems to use a Clevo L140CU Mainboard.

The main symptom is:

  iwlwifi 0000:02:00.0: Unable to change power state from D3hot to D0, device inaccessible
  nvme 0000:03:00.0: Unable to change power state from D3hot to D0, device inaccessible

and the machine is only partially usable after resume.  It can't run dmesg
and can't do a clean reboot.  This happens on every suspend/resume cycle.

Revert 5e85eba6f50d until we can figure out the root cause.

Fixes: 5e85eba6f50d ("PCI/ASPM: Refactor L1 PM Substates Control Register programming")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216877
Reported-by: Thomas Witt <kernel@witt.link>
Tested-by: Thomas Witt <kernel@witt.link>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org # v6.1+
Cc: Vidya Sagar <vidyas@nvidia.com>
21 months agoRevert "PCI/ASPM: Save L1 PM Substates Capability for suspend/resume"
Bjorn Helgaas [Fri, 3 Feb 2023 19:57:29 +0000 (13:57 -0600)]
Revert "PCI/ASPM: Save L1 PM Substates Capability for suspend/resume"

This reverts commit 4ff116d0d5fd8a025604b0802d93a2d5f4e465d1.

Tasev Nikola and Mark Enriquez reported that resume from suspend was broken
in v6.1-rc1.  Tasev bisected to a47126ec29f5 ("PCI/PTM: Cache PTM
Capability offset"), but we can't figure out how that could be related.

Mark saw the same symptoms and bisected to 4ff116d0d5fd ("PCI/ASPM: Save L1
PM Substates Capability for suspend/resume"), which does have a connection:
it restores L1 Substates configuration while ASPM L1 may be enabled:

  pci_restore_state
    pci_restore_aspm_l1ss_state
      aspm_program_l1ss
        pci_write_config_dword(PCI_L1SS_CTL1, ctl1)         # L1SS restore
    pci_restore_pcie_state
      pcie_capability_write_word(PCI_EXP_LNKCTL, cap[i++])  # L1 restore

which is a problem because PCIe r6.0, sec 5.5.4, requires that:

  If setting either or both of the enable bits for ASPM L1 PM
  Substates, both ports must be configured as described in this
  section while ASPM L1 is disabled.

Separately, Thomas Witt reported that 5e85eba6f50d ("PCI/ASPM: Refactor L1
PM Substates Control Register programming") broke suspend/resume, and it
depends on 4ff116d0d5fd.

Revert 4ff116d0d5fd ("PCI/ASPM: Save L1 PM Substates Capability for
suspend/resume") to fix the resume issue and enable revert of 5e85eba6f50d
to fix the issue Thomas reported.

Note that reverting 4ff116d0d5fd means L1 Substates config may be lost on
suspend/resume.  As far as we know the system will use more power but will
still *work* correctly.

Fixes: 4ff116d0d5fd ("PCI/ASPM: Save L1 PM Substates Capability for suspend/resume")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216782
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216877
Reported-by: Tasev Nikola <tasev.stefanoska@skynet.be>
Reported-by: Mark Enriquez <enriquezmark36@gmail.com>
Reported-by: Thomas Witt <kernel@witt.link>
Tested-by: Mark Enriquez <enriquezmark36@gmail.com>
Tested-by: Thomas Witt <kernel@witt.link>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org # v6.1+
Cc: Vidya Sagar <vidyas@nvidia.com>
21 months agoMerge tag 'soc-fixes-6.2-4' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc
Linus Torvalds [Fri, 10 Feb 2023 17:48:42 +0000 (09:48 -0800)]
Merge tag 'soc-fixes-6.2-4' of git://git./linux/kernel/git/soc/soc

Pull ARM SoC fixes from Arnd Bergmann:
 "All the changes this time are minor devicetree corrections, the
  majority being for 64-bit Rockchip SoC support. These are a couple of
  corrections for properties that are in violation of the binding, some
  that put the machine into safer operating points for the eMMC and
  thermal settings, and missing properties that prevented rk356x PCIe
  and ethernet from working correctly.

  The changes for amlogic and mediatek address incorrect properties that
  were preventing the display support on MT8195 and the MMC support on
  various Meson SoCs from working correctly.

  The stihxxx-b2120 change fixes the GPIO polarity for the DVB tuner to
  allow this to be used correctly after a futre driver change, though it
  has no effect on older kernels"

* tag 'soc-fixes-6.2-4' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc:
  arm64: dts: meson-gx: Make mmc host controller interrupts level-sensitive
  arm64: dts: meson-g12-common: Make mmc host controller interrupts level-sensitive
  arm64: dts: meson-axg: Make mmc host controller interrupts level-sensitive
  ARM: dts: stihxxx-b2120: fix polarity of reset line of tsin0 port
  arm64: dts: mediatek: mt8195: Fix vdosys* compatible strings
  arm64: dts: rockchip: align rk3399 DMC OPP table with bindings
  arm64: dts: rockchip: set sdmmc0 speed to sd-uhs-sdr50 on rock-3a
  arm64: dts: rockchip: fix probe of analog sound card on rock-3a
  arm64: dts: rockchip: add missing #interrupt-cells to rk356x pcie2x1
  arm64: dts: rockchip: fix input enable pinconf on rk3399
  ARM: dts: rockchip: add power-domains property to dp node on rk3288
  arm64: dts: rockchip: add io domain setting to rk3566-box-demo
  arm64: dts: rockchip: remove unsupported property from sdmmc2 for rock-3a
  arm64: dts: rockchip: drop unused LED mode property from rk3328-roc-cc
  arm64: dts: rockchip: reduce thermal limits on rk3399-pinephone-pro
  arm64: dts: rockchip: use correct reset names for rk3399 crypto nodes

21 months agoMerge tag 'riscv-for-linus-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 10 Feb 2023 17:27:52 +0000 (09:27 -0800)]
Merge tag 'riscv-for-linus-6.2-rc8' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Palmer Dabbelt:
 "This is a little bigger that I'd hope for this late in the cycle, but
  they're all pretty concrete fixes and the only one that's bigger than
  a few lines is pmdp_collapse_flush() (which is almost all
  boilerplate/comment). It's also all bug fixes for issues that have
  been around for a while.

  So I think it's not all that scary, just bad timing.

   - avoid partial TLB fences for huge pages, which are disallowed by
     the ISA

   - avoid missing a frame when dumping stacks

   - avoid misaligned accesses (and possibly overflows) in kprobes

   - fix a race condition in tracking page dirtiness"

* tag 'riscv-for-linus-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: Fixup race condition on PG_dcache_clean in flush_icache_pte
  riscv: kprobe: Fixup misaligned load text
  riscv: stacktrace: Fix missing the first frame
  riscv: mm: Implement pmdp_collapse_flush for THP

21 months agoMerge tag 'ceph-for-6.2-rc8' of https://github.com/ceph/ceph-client
Linus Torvalds [Fri, 10 Feb 2023 17:04:00 +0000 (09:04 -0800)]
Merge tag 'ceph-for-6.2-rc8' of https://github.com/ceph/ceph-client

Pull ceph fix from Ilya Dryomov:
 "A fix for a pretty embarrassing omission in the session flush handler
  from Xiubo, marked for stable"

* tag 'ceph-for-6.2-rc8' of https://github.com/ceph/ceph-client:
  ceph: flush cap releases when the session is flushed

21 months agoMerge tag 'block-6.2-2023-02-10' of git://git.kernel.dk/linux
Linus Torvalds [Fri, 10 Feb 2023 16:55:09 +0000 (08:55 -0800)]
Merge tag 'block-6.2-2023-02-10' of git://git.kernel.dk/linux

Pull block fix from Jens Axboe:
 "A single fix for a smatch regression introduced in this merge window"

* tag 'block-6.2-2023-02-10' of git://git.kernel.dk/linux:
  nvme-auth: mark nvme_auth_wq static

21 months agoMerge tag 'sound-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 10 Feb 2023 16:37:48 +0000 (08:37 -0800)]
Merge tag 'sound-6.2-rc8' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Hopefully the last one for 6.2, a collection of the fixes that have
  been gathered since the last pull.

  All changes are small and trivial device-specific fixes"

* tag 'sound-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda/realtek: Add Positivo N14KP6-TG
  ASoC: topology: Return -ENOMEM on memory allocation failure
  ALSA: emux: Avoid potential array out-of-bound in snd_emux_xg_control()
  ASoC: fsl_sai: fix getting version from VERID
  ALSA: hda/realtek: fix mute/micmute LEDs don't work for a HP platform.
  ALSA: hda/realtek: Add quirk for ASUS UM3402 using CS35L41
  ASoC: codecs: es8326: Fix DTS properties reading
  ASoC: tas5805m: add missing page switch.
  ASoC: tas5805m: rework to avoid scheduling while atomic.
  ALSA: hda/realtek: Enable mute/micmute LEDs on HP Elitebook, 645 G9
  ASoC: SOF: amd: Fix for handling spurious interrupts from DSP
  ALSA: hda/realtek: Fix the speaker output on Samsung Galaxy Book2 Pro 360
  ALSA: pci: lx6464es: fix a debug loop
  ASoC: rt715-sdca: fix clock stop prepare timeout issue

21 months agoMerge tag 'amlogic-fixes-v6.2-rc-take2' of https://git.kernel.org/pub/scm/linux/kerne...
Arnd Bergmann [Fri, 10 Feb 2023 10:31:24 +0000 (11:31 +0100)]
Merge tag 'amlogic-fixes-v6.2-rc-take2' of https://git./linux/kernel/git/amlogic/linux into arm/fixes

Amlogic fixes for v6.2-rc, take2:
- Change MMC controllers interrupts flag to level on all families, fixes irq loss & performance issues when cpu loaded

* tag 'amlogic-fixes-v6.2-rc-take2' of https://git.kernel.org/pub/scm/linux/kernel/git/amlogic/linux:
  arm64: dts: meson-gx: Make mmc host controller interrupts level-sensitive
  arm64: dts: meson-g12-common: Make mmc host controller interrupts level-sensitive
  arm64: dts: meson-axg: Make mmc host controller interrupts level-sensitive

Link: https://lore.kernel.org/r/761c2ebc-7c93-8504-35ae-3e84ad216bcf@linaro.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
21 months agoarm64: dts: meson-gx: Make mmc host controller interrupts level-sensitive
Heiner Kallweit [Thu, 9 Feb 2023 20:11:47 +0000 (21:11 +0100)]
arm64: dts: meson-gx: Make mmc host controller interrupts level-sensitive

The usage of edge-triggered interrupts lead to lost interrupts under load,
see [0]. This was confirmed to be fixed by using level-triggered
interrupts.
The report was about SDIO. However, as the host controller is the same
for SD and MMC, apply the change to all mmc controller instances.

[0] https://www.spinics.net/lists/linux-mmc/msg73991.html

Fixes: ef8d2ffedf18 ("ARM64: dts: meson-gxbb: add MMC support")
Cc: stable@vger.kernel.org
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Acked-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/76e042e0-a610-5ed5-209f-c4d7f879df44@gmail.com
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
21 months agoarm64: dts: meson-g12-common: Make mmc host controller interrupts level-sensitive
Heiner Kallweit [Thu, 9 Feb 2023 20:11:10 +0000 (21:11 +0100)]
arm64: dts: meson-g12-common: Make mmc host controller interrupts level-sensitive

The usage of edge-triggered interrupts lead to lost interrupts under load,
see [0]. This was confirmed to be fixed by using level-triggered
interrupts.
The report was about SDIO. However, as the host controller is the same
for SD and MMC, apply the change to all mmc controller instances.

[0] https://www.spinics.net/lists/linux-mmc/msg73991.html

Fixes: 4759fd87b928 ("arm64: dts: meson: g12a: add mmc nodes")
Tested-by: FUKAUMI Naoki <naoki@radxa.com>
Tested-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Tested-by: Jerome Brunet <jbrunet@baylibre.com>
Cc: stable@vger.kernel.org
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Acked-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/27d89baa-b8fa-baca-541b-ef17a97cde3c@gmail.com
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
21 months agoarm64: dts: meson-axg: Make mmc host controller interrupts level-sensitive
Heiner Kallweit [Thu, 9 Feb 2023 20:10:31 +0000 (21:10 +0100)]
arm64: dts: meson-axg: Make mmc host controller interrupts level-sensitive

The usage of edge-triggered interrupts lead to lost interrupts under load,
see [0]. This was confirmed to be fixed by using level-triggered
interrupts.
The report was about SDIO. However, as the host controller is the same
for SD and MMC, apply the change to all mmc controller instances.

[0] https://www.spinics.net/lists/linux-mmc/msg73991.html

Fixes: 221cf34bac54 ("ARM64: dts: meson-axg: enable the eMMC controller")
Reported-by: Peter Suti <peter.suti@streamunlimited.com>
Tested-by: Vyacheslav Bocharov <adeep@lexina.in>
Tested-by: Peter Suti <peter.suti@streamunlimited.com>
Cc: stable@vger.kernel.org
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Acked-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/c00655d3-02f8-6f5f-4239-ca2412420cad@gmail.com
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
21 months agoMerge tag 'drm-fixes-2023-02-10' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 10 Feb 2023 01:52:00 +0000 (17:52 -0800)]
Merge tag 'drm-fixes-2023-02-10' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Weekly fixes.

  The amdgpu had a few small fixes to display flicker on certain
  configurations, however it was found the the flicker was lessened but
  there were other unintended consequences, so for now they've been
  reverted and replaced with an option for users to test with so future
  fixes can be developed.

  Otherwise apart from the usual bunch of i915 and amdgpu, there's a
  client, virtio-gpu and an nvidiafb fix that reorders its loading to
  avoid failure.

  client:
   - refcount fix

  amdgpu:
   - a bunch of attempted flicker fixes that regressed turned into a
     user workaround option for now
   - Properly fix S/G display with AGP aperture enabled
   - Fix cursor offset with 180 rotation
   - SMU13 fixes
   - Use TGID for GPUVM traces
   - Fix oops on in fence error path
   - Don't run IB tests on hw rings when sw rings are in use
   - memory leak fix

  i915:
   - Display watermark fix
   - fbdev fix for PSR, FBC, DRRS
   - Move fd_install after last use of fence
   - Initialize the obj flags for shmem objects
   - Fix VBT DSI DVO port handling

  virtio-gpu:
   - fence fix

  nvidiafb:
   - regression fix for driver load when no hw supported"

* tag 'drm-fixes-2023-02-10' of git://anongit.freedesktop.org/drm/drm: (27 commits)
  Revert "drm/amd/display: disable S/G display on DCN 3.1.5"
  Revert "drm/amd/display: disable S/G display on DCN 2.1.0"
  Revert "drm/amd/display: disable S/G display on DCN 3.1.2/3"
  drm/amdgpu: add S/G display parameter
  drm/amdgpu/smu: skip pptable init under sriov
  amd/amdgpu: remove test ib on hw ring
  drm/amdgpu/fence: Fix oops due to non-matching drm_sched init/fini
  drm/amdgpu: Use the TGID for trace_amdgpu_vm_update_ptes
  drm/amdgpu: Add unique_id support for GC 11.0.1/2
  drm/amd/pm: bump SMU 13.0.7 driver_if header version
  drm/amd/pm: bump SMU 13.0.0 driver_if header version
  drm/amd/pm: add SMU 13.0.7 missing GetPptLimit message mapping
  drm/amd/display: fix cursor offset on rotation 180
  drm/amd/amdgpu: enable athub cg 11.0.3
  Revert "drm/amd/display: disable S/G display on DCN 3.1.4"
  drm/amd/display: properly handling AGP aperture in vm setup
  drm/amd/display: disable S/G display on DCN 3.1.2/3
  drm/amd/display: disable S/G display on DCN 2.1.0
  drm/i915: Fix VBT DSI DVO port handling
  drm/client: fix circular reference counting issue
  ...

21 months agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Fri, 10 Feb 2023 01:34:14 +0000 (17:34 -0800)]
Merge tag 'for-linus' of git://git./linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "The usual collection of small driver bug fixes:

   - Fix error unwind bugs in hfi1, irdma rtrs

   - Old bug with IPoIB children interfaces possibly using the wrong
     number of queues

   - Really old bug in usnic calling iommu_map in an atomic context

   - Recent regression from the DMABUF locking rework

   - Missing user data validation in MANA"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/rtrs: Don't call kobject_del for srv_path->kobj
  RDMA/mana_ib: Prevent array underflow in mana_ib_create_qp_raw()
  IB/hfi1: Assign npages earlier
  RDMA/umem: Use dma-buf locked API to solve deadlock
  RDMA/usnic: use iommu_map_atomic() under spin_lock()
  RDMA/irdma: Fix potential NULL-ptr-dereference
  IB/IPoIB: Fix legacy IPoIB due to wrong number of queues
  IB/hfi1: Restore allocated resources on failed copyout

21 months agoMerge tag 'amd-drm-fixes-6.2-2023-02-09' of https://gitlab.freedesktop.org/agd5f...
Dave Airlie [Thu, 9 Feb 2023 23:49:12 +0000 (09:49 +1000)]
Merge tag 'amd-drm-fixes-6.2-2023-02-09' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-6.2-2023-02-09:

amdgpu:
- Add a parameter to disable S/G display
- Re-enable S/G display on all DCNs

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230209174504.7577-1-alexander.deucher@amd.com
21 months agoMerge tag 'drm-intel-fixes-2023-02-09' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Thu, 9 Feb 2023 23:47:20 +0000 (09:47 +1000)]
Merge tag 'drm-intel-fixes-2023-02-09' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

- Display watermark fix (Ville)
- fbdev fix for PSR, FBC, DRRS (Jouni)
- Move fd_install after last use of fence (Rob)
- Initialize the obj flags for shmem objects (Aravind)
- Fix VBT DSI DVO port handling (Ville)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/Y+UZ0rh2YlhTrE4t@intel.com
21 months agoMerge tag 'drm-misc-fixes-2023-02-09' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Thu, 9 Feb 2023 23:15:52 +0000 (09:15 +1000)]
Merge tag 'drm-misc-fixes-2023-02-09' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

A fix for a circular refcounting in drm/client, one for a memory leak in
amdgpu and a virtio fence fix when interrupted

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20230209083600.7hi6roht6xxgldgz@houat
21 months agoriscv: Fixup race condition on PG_dcache_clean in flush_icache_pte
Guo Ren [Fri, 27 Jan 2023 03:53:06 +0000 (22:53 -0500)]
riscv: Fixup race condition on PG_dcache_clean in flush_icache_pte

In commit 588a513d3425 ("arm64: Fix race condition on PG_dcache_clean
in __sync_icache_dcache()"), we found RISC-V has the same issue as the
previous arm64. The previous implementation didn't guarantee the correct
sequence of operations, which means flush_icache_all() hasn't been
called when the PG_dcache_clean was set. That would cause a risk of page
synchronization.

Fixes: 08f051eda33b ("RISC-V: Flush I$ when making a dirty page executable")
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@kernel.org>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230127035306.1819561-1-guoren@kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
21 months agoriscv: kprobe: Fixup misaligned load text
Guo Ren [Sat, 4 Feb 2023 06:35:31 +0000 (01:35 -0500)]
riscv: kprobe: Fixup misaligned load text

The current kprobe would cause a misaligned load for the probe point.
This patch fixup it with two half-word loads instead.

Fixes: c22b0bcb1dd0 ("riscv: Add kprobes supported")
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@kernel.org>
Link: https://lore.kernel.org/linux-riscv/878rhig9zj.fsf@all.your.base.are.belong.to.us/
Reported-by: Bjorn Topel <bjorn.topel@gmail.com>
Reviewed-by: Björn Töpel <bjorn@kernel.org>
Link: https://lore.kernel.org/r/20230204063531.740220-1-guoren@kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
21 months agoMerge tag 'pm-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Thu, 9 Feb 2023 18:54:57 +0000 (10:54 -0800)]
Merge tag 'pm-6.2-rc8' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fix from Rafael Wysocki:
 "Fix the incorrect value returned by cpufreq driver's ->get() callback
  for Qualcomm platforms (Douglas Anderson)"

* tag 'pm-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  cpufreq: qcom-hw: Fix cpufreq_driver->get() for non-LMH systems

21 months agoMerge tag 'net-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 9 Feb 2023 17:17:38 +0000 (09:17 -0800)]
Merge tag 'net-6.2-rc8' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Paolo Abeni:
 "Including fixes from can and ipsec subtrees.

  Current release - regressions:

   - sched: fix off by one in htb_activate_prios()

   - eth: mana: fix accessing freed irq affinity_hint

   - eth: ice: fix out-of-bounds KASAN warning in virtchnl

  Current release - new code bugs:

   - eth: mtk_eth_soc: enable special tag when any MAC uses DSA

  Previous releases - always broken:

   - core: fix sk->sk_txrehash default

   - neigh: make sure used and confirmed times are valid

   - mptcp: be careful on subflow status propagation on errors

   - xfrm: prevent potential spectre v1 gadget in xfrm_xlate32_attr()

   - phylink: move phy_device_free() to correctly release phy device

   - eth: mlx5:
      - fix crash unsetting rx-vlan-filter in switchdev mode
      - fix hang on firmware reset
      - serialize module cleanup with reload and remove"

* tag 'net-6.2-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (57 commits)
  selftests: forwarding: lib: quote the sysctl values
  net: mscc: ocelot: fix all IPv6 getting trapped to CPU when PTP timestamping is used
  rds: rds_rm_zerocopy_callback() use list_first_entry()
  net: txgbe: Update support email address
  selftests: Fix failing VXLAN VNI filtering test
  selftests: mptcp: stop tests earlier
  selftests: mptcp: allow more slack for slow test-case
  mptcp: be careful on subflow status propagation on errors
  mptcp: fix locking for in-kernel listener creation
  mptcp: fix locking for setsockopt corner-case
  mptcp: do not wait for bare sockets' timeout
  net: ethernet: mtk_eth_soc: fix DSA TX tag hwaccel for switch port 0
  nfp: ethtool: fix the bug of setting unsupported port speed
  txhash: fix sk->sk_txrehash default
  net: ethernet: mtk_eth_soc: fix wrong parameters order in __xdp_rxq_info_reg()
  net: ethernet: mtk_eth_soc: enable special tag when any MAC uses DSA
  net: sched: sch: Fix off by one in htb_activate_prios()
  igc: Add ndo_tx_timeout support
  net: mana: Fix accessing freed irq affinity_hint
  hv_netvsc: Allocate memory in netvsc_dma_map() with GFP_ATOMIC
  ...

21 months agoMerge tag 'for-linus-2023020901' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 9 Feb 2023 17:09:13 +0000 (09:09 -0800)]
Merge tag 'for-linus-2023020901' of git://git./linux/kernel/git/hid/hid

Pull HID fixes from Benjamin Tissoires:

 - fix potential infinite loop with a badly crafted HID device (Xin
   Zhao)

 - fix regression from 6.1 in USB logitech devices potentially making
   their mouse wheel not working (Bastien Nocera)

 - clean up in AMD sensors, which fixes a long time resume bug (Mario
   Limonciello)

 - few device small fixes and quirks

* tag 'for-linus-2023020901' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid:
  HID: Ignore battery for ELAN touchscreen 29DF on HP
  HID: amd_sfh: if no sensors are enabled, clean up
  HID: logitech: Disable hi-res scrolling on USB
  HID: core: Fix deadloop in hid_apply_multiplier.
  HID: Ignore battery for Elan touchscreen on Asus TP420IA
  HID: elecom: add support for TrackBall 056E:011C

21 months agoMerge tag '6.2-rc8-smb3-client-fix' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Thu, 9 Feb 2023 17:00:26 +0000 (09:00 -0800)]
Merge tag '6.2-rc8-smb3-client-fix' of git://git.samba.org/sfrench/cifs-2.6

Pull cifx fix from Steve French:
 "Small fix for use after free"

* tag '6.2-rc8-smb3-client-fix' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: Fix use-after-free in rdata->read_into_pages()

21 months agobtrfs: free device in btrfs_close_devices for a single device filesystem
Anand Jain [Fri, 20 Jan 2023 13:47:16 +0000 (21:47 +0800)]
btrfs: free device in btrfs_close_devices for a single device filesystem

We have this check to make sure we don't accidentally add older devices
that may have disappeared and re-appeared with an older generation from
being added to an fs_devices (such as a replace source device). This
makes sense, we don't want stale disks in our file system. However for
single disks this doesn't really make sense.

I've seen this in testing, but I was provided a reproducer from a
project that builds btrfs images on loopback devices. The loopback
device gets cached with the new generation, and then if it is re-used to
generate a new file system we'll fail to mount it because the new fs is
"older" than what we have in cache.

Fix this by freeing the cache when closing the device for a single device
filesystem. This will ensure that the mount command passed device path is
scanned successfully during the next mount.

CC: stable@vger.kernel.org # 5.10+
Reported-by: Daan De Meyer <daandemeyer@fb.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
21 months agobtrfs: lock the inode in shared mode before starting fiemap
Filipe Manana [Mon, 23 Jan 2023 16:54:46 +0000 (16:54 +0000)]
btrfs: lock the inode in shared mode before starting fiemap

Currently fiemap does not take the inode's lock (VFS lock), it only locks
a file range in the inode's io tree. This however can lead to a deadlock
if we have a concurrent fsync on the file and fiemap code triggers a fault
when accessing the user space buffer with fiemap_fill_next_extent(). The
deadlock happens on the inode's i_mmap_lock semaphore, which is taken both
by fsync and btrfs_page_mkwrite(). This deadlock was recently reported by
syzbot and triggers a trace like the following:

   task:syz-executor361 state:D stack:20264 pid:5668  ppid:5119   flags:0x00004004
   Call Trace:
    <TASK>
    context_switch kernel/sched/core.c:5293 [inline]
    __schedule+0x995/0xe20 kernel/sched/core.c:6606
    schedule+0xcb/0x190 kernel/sched/core.c:6682
    wait_on_state fs/btrfs/extent-io-tree.c:707 [inline]
    wait_extent_bit+0x577/0x6f0 fs/btrfs/extent-io-tree.c:751
    lock_extent+0x1c2/0x280 fs/btrfs/extent-io-tree.c:1742
    find_lock_delalloc_range+0x4e6/0x9c0 fs/btrfs/extent_io.c:488
    writepage_delalloc+0x1ef/0x540 fs/btrfs/extent_io.c:1863
    __extent_writepage+0x736/0x14e0 fs/btrfs/extent_io.c:2174
    extent_write_cache_pages+0x983/0x1220 fs/btrfs/extent_io.c:3091
    extent_writepages+0x219/0x540 fs/btrfs/extent_io.c:3211
    do_writepages+0x3c3/0x680 mm/page-writeback.c:2581
    filemap_fdatawrite_wbc+0x11e/0x170 mm/filemap.c:388
    __filemap_fdatawrite_range mm/filemap.c:421 [inline]
    filemap_fdatawrite_range+0x175/0x200 mm/filemap.c:439
    btrfs_fdatawrite_range fs/btrfs/file.c:3850 [inline]
    start_ordered_ops fs/btrfs/file.c:1737 [inline]
    btrfs_sync_file+0x4ff/0x1190 fs/btrfs/file.c:1839
    generic_write_sync include/linux/fs.h:2885 [inline]
    btrfs_do_write_iter+0xcd3/0x1280 fs/btrfs/file.c:1684
    call_write_iter include/linux/fs.h:2189 [inline]
    new_sync_write fs/read_write.c:491 [inline]
    vfs_write+0x7dc/0xc50 fs/read_write.c:584
    ksys_write+0x177/0x2a0 fs/read_write.c:637
    do_syscall_x64 arch/x86/entry/common.c:50 [inline]
    do_syscall_64+0x3d/0xb0 arch/x86/entry/common.c:80
    entry_SYSCALL_64_after_hwframe+0x63/0xcd
   RIP: 0033:0x7f7d4054e9b9
   RSP: 002b:00007f7d404fa2f8 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
   RAX: ffffffffffffffda RBX: 00007f7d405d87a0 RCX: 00007f7d4054e9b9
   RDX: 0000000000000090 RSI: 0000000020000000 RDI: 0000000000000006
   RBP: 00007f7d405a51d0 R08: 0000000000000000 R09: 0000000000000000
   R10: 0000000000000000 R11: 0000000000000246 R12: 61635f65646f6e69
   R13: 65646f7475616f6e R14: 7261637369646f6e R15: 00007f7d405d87a8
    </TASK>
   INFO: task syz-executor361:5697 blocked for more than 145 seconds.
         Not tainted 6.2.0-rc3-syzkaller-00376-g7c6984405241 #0
   "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
   task:syz-executor361 state:D stack:21216 pid:5697  ppid:5119   flags:0x00004004
   Call Trace:
    <TASK>
    context_switch kernel/sched/core.c:5293 [inline]
    __schedule+0x995/0xe20 kernel/sched/core.c:6606
    schedule+0xcb/0x190 kernel/sched/core.c:6682
    rwsem_down_read_slowpath+0x5f9/0x930 kernel/locking/rwsem.c:1095
    __down_read_common+0x54/0x2a0 kernel/locking/rwsem.c:1260
    btrfs_page_mkwrite+0x417/0xc80 fs/btrfs/inode.c:8526
    do_page_mkwrite+0x19e/0x5e0 mm/memory.c:2947
    wp_page_shared+0x15e/0x380 mm/memory.c:3295
    handle_pte_fault mm/memory.c:4949 [inline]
    __handle_mm_fault mm/memory.c:5073 [inline]
    handle_mm_fault+0x1b79/0x26b0 mm/memory.c:5219
    do_user_addr_fault+0x69b/0xcb0 arch/x86/mm/fault.c:1428
    handle_page_fault arch/x86/mm/fault.c:1519 [inline]
    exc_page_fault+0x7a/0x110 arch/x86/mm/fault.c:1575
    asm_exc_page_fault+0x22/0x30 arch/x86/include/asm/idtentry.h:570
   RIP: 0010:copy_user_short_string+0xd/0x40 arch/x86/lib/copy_user_64.S:233
   Code: 74 0a 89 (...)
   RSP: 0018:ffffc9000570f330 EFLAGS: 00050202
   RAX: ffffffff843e6601 RBX: 00007fffffffefc8 RCX: 0000000000000007
   RDX: 0000000000000000 RSI: ffffc9000570f3e0 RDI: 0000000020000120
   RBP: ffffc9000570f490 R08: 0000000000000000 R09: fffff52000ae1e83
   R10: fffff52000ae1e83 R11: 1ffff92000ae1e7c R12: 0000000000000038
   R13: ffffc9000570f3e0 R14: 0000000020000120 R15: ffffc9000570f3e0
    copy_user_generic arch/x86/include/asm/uaccess_64.h:37 [inline]
    raw_copy_to_user arch/x86/include/asm/uaccess_64.h:58 [inline]
    _copy_to_user+0xe9/0x130 lib/usercopy.c:34
    copy_to_user include/linux/uaccess.h:169 [inline]
    fiemap_fill_next_extent+0x22e/0x410 fs/ioctl.c:144
    emit_fiemap_extent+0x22d/0x3c0 fs/btrfs/extent_io.c:3458
    fiemap_process_hole+0xa00/0xad0 fs/btrfs/extent_io.c:3716
    extent_fiemap+0xe27/0x2100 fs/btrfs/extent_io.c:3922
    btrfs_fiemap+0x172/0x1e0 fs/btrfs/inode.c:8209
    ioctl_fiemap fs/ioctl.c:219 [inline]
    do_vfs_ioctl+0x185b/0x2980 fs/ioctl.c:810
    __do_sys_ioctl fs/ioctl.c:868 [inline]
    __se_sys_ioctl+0x83/0x170 fs/ioctl.c:856
    do_syscall_x64 arch/x86/entry/common.c:50 [inline]
    do_syscall_64+0x3d/0xb0 arch/x86/entry/common.c:80
    entry_SYSCALL_64_after_hwframe+0x63/0xcd
   RIP: 0033:0x7f7d4054e9b9
   RSP: 002b:00007f7d390d92f8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
   RAX: ffffffffffffffda RBX: 00007f7d405d87b0 RCX: 00007f7d4054e9b9
   RDX: 0000000020000100 RSI: 00000000c020660b RDI: 0000000000000005
   RBP: 00007f7d405a51d0 R08: 00007f7d390d9700 R09: 0000000000000000
   R10: 00007f7d390d9700 R11: 0000000000000246 R12: 61635f65646f6e69
   R13: 65646f7475616f6e R14: 7261637369646f6e R15: 00007f7d405d87b8
    </TASK>

What happens is the following:

1) Task A is doing an fsync, enters btrfs_sync_file() and flushes delalloc
   before locking the inode and the i_mmap_lock semaphore, that is, before
   calling btrfs_inode_lock();

2) After task A flushes delalloc and before it calls btrfs_inode_lock(),
   another task dirties a page;

3) Task B starts a fiemap without FIEMAP_FLAG_SYNC, so the page dirtied
   at step 2 remains dirty and unflushed. Then when it enters
   extent_fiemap() and it locks a file range that includes the range of
   the page dirtied in step 2;

4) Task A calls btrfs_inode_lock() and locks the inode (VFS lock) and the
   inode's i_mmap_lock semaphore in write mode. Then it tries to flush
   delalloc by calling start_ordered_ops(), which will block, at
   find_lock_delalloc_range(), when trying to lock the range of the page
   dirtied at step 2, since this range was locked by the fiemap task (at
   step 3);

5) Task B generates a page fault when accessing the user space fiemap
   buffer with a call to fiemap_fill_next_extent().

   The fault handler needs to call btrfs_page_mkwrite() for some other
   page of our inode, and there we deadlock when trying to lock the
   inode's i_mmap_lock semaphore in read mode, since the fsync task locked
   it in write mode (step 4) and the fsync task can not progress because
   it's waiting to lock a file range that is currently locked by us (the
   fiemap task, step 3).

Fix this by taking the inode's lock (VFS lock) in shared mode when
entering fiemap. This effectively serializes fiemap with fsync (except the
most expensive part of fsync, the log sync), preventing this deadlock.

Reported-by: syzbot+cc35f55c41e34c30dcb5@syzkaller.appspotmail.com
Link: https://lore.kernel.org/linux-btrfs/00000000000032dc7305f2a66f46@google.com/
CC: stable@vger.kernel.org # 6.1+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
21 months agoRevert "drm/amd/display: disable S/G display on DCN 3.1.5"
Alex Deucher [Thu, 9 Feb 2023 14:09:45 +0000 (09:09 -0500)]
Revert "drm/amd/display: disable S/G display on DCN 3.1.5"

This reverts commit 3cc67fe1b3aa1ac4720e002f2aa2d08c9199a584.

Some users have reported flickerng with S/G display.  We've
tried extensively to reproduce and debug the issue on a wide
variety of platform configurations (DRAM bandwidth, etc.) and
a variety of monitors, but so far have not been able to.  We
disabled S/G display on a number of platforms to address this
but that leads to failure to pin framebuffers errors and
blank displays when there is memory pressure or no displays
at all on systems with limited carveout (e.g., Chromebooks).
We have a parameter to disable this as a debugging option as a
way for users to disable this, depending on their use case,
and for us to help debug this further.  Having this enabled
seems like the lesser of to evils.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
21 months agoRevert "drm/amd/display: disable S/G display on DCN 2.1.0"
Alex Deucher [Thu, 9 Feb 2023 14:09:19 +0000 (09:09 -0500)]
Revert "drm/amd/display: disable S/G display on DCN 2.1.0"

This reverts commit 2404f9b0ea0153c3fddb0c4d7a43869dc8608f6f.

Some users have reported flickerng with S/G display.  We've
tried extensively to reproduce and debug the issue on a wide
variety of platform configurations (DRAM bandwidth, etc.) and
a variety of monitors, but so far have not been able to.  We
disabled S/G display on a number of platforms to address this
but that leads to failure to pin framebuffers errors and
blank displays when there is memory pressure or no displays
at all on systems with limited carveout (e.g., Chromebooks).
We have a parameter to disable this as a debugging option as a
way for users to disable this, depending on their use case,
and for us to help debug this further.  Having this enabled
seems like the lesser of to evils.

Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>