From: Zhenguo Yao <yaozhenguo1@gmail.com>
To: bhelgaas@google.com, alex.williamson@redhat.com
Cc: cohuck@redhat.com, jgg@ziepe.ca, mgurtovoy@nvidia.com,
yishaih@nvidia.com, kvm@vger.kernel.org,
linux-pci@vger.kernel.org, linux-kernel@vger.kernel.org,
yaozhenguo@jd.com, Zhenguo Yao <yaozhenguo1@gmail.com>
Subject: [PATCH v1 2/2] vfio-pci: Don't do device reset when ignore_reset is setting
Date: Thu, 14 Oct 2021 17:57:48 +0800 [thread overview]
Message-ID: <20211014095748.84604-3-yaozhenguo1@gmail.com> (raw)
In-Reply-To: <20211014095748.84604-1-yaozhenguo1@gmail.com>
In some scenarios, vfio device can't do any reset in initialization
process. For example: Nvswitch and GPU A100 working in Shared NVSwitch
Virtualization Model. In such mode, The GPUs can't do any reset when
Guest VM is booting up.
So, Using ignore_reset to control whether to do PCI reset in
initialization. In Shared NVSwitch Virtualization Model, GPUs will
ignore reset when Gust VM booting up.
Signed-off-by: Zhenguo Yao <yaozhenguo1@gmail.com>
---
drivers/vfio/pci/vfio_pci_core.c | 48 ++++++++++++++++++++------------
1 file changed, 30 insertions(+), 18 deletions(-)
diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
index 68198e0f2a63..83d3ef5d3a9c 100644
--- a/drivers/vfio/pci/vfio_pci_core.c
+++ b/drivers/vfio/pci/vfio_pci_core.c
@@ -254,11 +254,13 @@ int vfio_pci_core_enable(struct vfio_pci_core_device *vdev)
if (ret)
return ret;
- /* If reset fails because of the device lock, fail this path entirely */
- ret = pci_try_reset_function(pdev);
- if (ret == -EAGAIN) {
- pci_disable_device(pdev);
- return ret;
+ if (!pdev->ignore_reset) {
+ /* If reset fails because of the device lock, fail this path entirely */
+ ret = pci_try_reset_function(pdev);
+ if (ret == -EAGAIN) {
+ pci_disable_device(pdev);
+ return ret;
+ }
}
vdev->reset_works = !ret;
@@ -388,25 +390,30 @@ void vfio_pci_core_disable(struct vfio_pci_core_device *vdev)
*/
pci_write_config_word(pdev, PCI_COMMAND, PCI_COMMAND_INTX_DISABLE);
- /*
- * Try to get the locks ourselves to prevent a deadlock. The
- * success of this is dependent on being able to lock the device,
- * which is not always possible.
- * We can not use the "try" reset interface here, which will
- * overwrite the previously restored configuration information.
- */
- if (vdev->reset_works && pci_dev_trylock(pdev)) {
- if (!__pci_reset_function_locked(pdev))
- vdev->needs_reset = false;
- pci_dev_unlock(pdev);
+ if (!pdev->ignore_reset) {
+ /*
+ * Try to get the locks ourselves to prevent a deadlock. The
+ * success of this is dependent on being able to lock the device,
+ * which is not always possible.
+ * We can not use the "try" reset interface here, which will
+ * overwrite the previously restored configuration information.
+ */
+ if (vdev->reset_works && pci_dev_trylock(pdev)) {
+ if (!__pci_reset_function_locked(pdev))
+ vdev->needs_reset = false;
+ pci_dev_unlock(pdev);
+ }
}
pci_restore_state(pdev);
out:
pci_disable_device(pdev);
- if (!vfio_pci_dev_set_try_reset(vdev->vdev.dev_set) && !disable_idle_d3)
- vfio_pci_set_power_state(vdev, PCI_D3hot);
+ if (!pdev->ignore_reset) {
+ if (!vfio_pci_dev_set_try_reset(vdev->vdev.dev_set) &&
+ !disable_idle_d3)
+ vfio_pci_set_power_state(vdev, PCI_D3hot);
+ }
}
EXPORT_SYMBOL_GPL(vfio_pci_core_disable);
@@ -919,6 +926,8 @@ long vfio_pci_core_ioctl(struct vfio_device *core_vdev, unsigned int cmd,
if (!vdev->reset_works)
return -EINVAL;
+ if (vdev->pdev->ignore_reset)
+ return -EINVAL;
vfio_pci_zap_and_down_write_memory_lock(vdev);
ret = pci_try_reset_function(vdev->pdev);
@@ -1007,6 +1016,9 @@ long vfio_pci_core_ioctl(struct vfio_device *core_vdev, unsigned int cmd,
bool slot = false;
int group_idx, count = 0, ret = 0;
+ if (vdev->pdev->ignore_reset)
+ return -EINVAL;
+
minsz = offsetofend(struct vfio_pci_hot_reset, count);
if (copy_from_user(&hdr, (void __user *)arg, minsz))
--
2.27.0
next prev parent reply other threads:[~2021-10-14 9:58 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-10-14 9:57 [PATCH v1 0/2] Add ablility of VFIO driver to ignore reset when device don't need it Zhenguo Yao
2021-10-14 9:57 ` [PATCH v1 1/2] PCI: Add ignore_reset sysfs interface to control whether to do device reset in PCI drivers Zhenguo Yao
2021-10-14 9:57 ` Zhenguo Yao [this message]
2021-10-14 12:48 ` [PATCH v1 0/2] Add ablility of VFIO driver to ignore reset when device don't need it Alex Williamson
2021-10-14 13:37 ` Zhenguo Yao
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20211014095748.84604-3-yaozhenguo1@gmail.com \
--to=yaozhenguo1@gmail.com \
--cc=alex.williamson@redhat.com \
--cc=bhelgaas@google.com \
--cc=cohuck@redhat.com \
--cc=jgg@ziepe.ca \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pci@vger.kernel.org \
--cc=mgurtovoy@nvidia.com \
--cc=yaozhenguo@jd.com \
--cc=yishaih@nvidia.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®