From: Bjorn Helgaas <helgaas@kernel.org>
To: Nicolin Chen <nicolinc@nvidia.com>
Cc: bhelgaas@google.com, joerg.roedel@amd.com, kevin.tian@intel.com,
jgg@nvidia.com, linux-pci@vger.kernel.org,
linux-kernel@vger.kernel.org, xueshuai@linux.alibaba.com
Subject: Re: [PATCH rc] PCI: Fix nested pci_dev_reset_iommu_prepare/done()
Date: Wed, 18 Mar 2026 17:57:05 -0500 [thread overview]
Message-ID: <20260318225705.GA479639@bhelgaas> (raw)
In-Reply-To: <20260318220028.1146905-1-nicolinc@nvidia.com>
On Wed, Mar 18, 2026 at 03:00:28PM -0700, Nicolin Chen wrote:
> Shuai found that cxl_reset_bus_function() calls pci_reset_bus_function()
> internally while both are calling pci_dev_reset_iommu_prepare/done().
>
> This results in a nested pci_dev_reset_iommu_prepare/done() call:
> cxl_reset_bus_function():
> pci_dev_reset_iommu_prepare(); // outer
> pci_reset_bus_function():
> pci_dev_reset_iommu_prepare(); // inner (re-entry)
> ...
> pci_dev_reset_iommu_done(); // inner
> pci_dev_reset_iommu_done(); // outer
>
> However, pci_dev_reset_iommu_prepare() doesn't allow a re-entry for now.
>
> Digging further, I found that most functions in pci_dev_specific_reset()
> except reset_ivb_igd() are calling pcie_flr() that has nested those two
> functions as well.
>
> Drop the outer callbacks in:
> - cxl_reset_bus_function()
> - pci_dev_specific_reset()
>
> Replace with adding the callbacks in:
> + reset_ivb_igd()
This needs to say clearly what the issue is. Deadlock? Crash?
Corruption?
I guess it's a deadlock when the second call to
pci_dev_reset_iommu_prepare() acquires the group->mutex?
Since f5b16b802174 appeared in v7.0-rc1 (merged by Joerg), I guess
this fix also needs to be in v7.0 -- please confirm.
> Fixes: f5b16b802174 ("PCI: Suspend iommu function prior to resetting a device")
> Cc: stable@vger.kernel.org
> Reported-by: Shuai Xue <xueshuai@linux.alibaba.com>
> Closes: https://lore.kernel.org/all/absKsk7qQOwzhpzv@Asurada-Nvidia/
> Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
> ---
> drivers/pci/pci.c | 7 -------
> drivers/pci/quirks.c | 29 +++++++++++------------------
> 2 files changed, 11 insertions(+), 25 deletions(-)
>
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 8479c2e1f74f..8f1fd083a84c 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -4958,12 +4958,6 @@ static int cxl_reset_bus_function(struct pci_dev *dev, bool probe)
> if (rc)
> return -ENOTTY;
>
> - rc = pci_dev_reset_iommu_prepare(dev);
> - if (rc) {
> - pci_err(dev, "failed to stop IOMMU for a PCI reset: %d\n", rc);
> - return rc;
> - }
> -
> if (reg & PCI_DVSEC_CXL_PORT_CTL_UNMASK_SBR) {
> val = reg;
> } else {
> @@ -4978,7 +4972,6 @@ static int cxl_reset_bus_function(struct pci_dev *dev, bool probe)
> pci_write_config_word(bridge, dvsec + PCI_DVSEC_CXL_PORT_CTL,
> reg);
>
> - pci_dev_reset_iommu_done(dev);
> return rc;
> }
>
> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> index 48946cca4be7..d1d210bc2a15 100644
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -3973,6 +3973,7 @@ static int reset_ivb_igd(struct pci_dev *dev, bool probe)
> void __iomem *mmio_base;
> unsigned long timeout;
> u32 val;
> + int ret;
>
> if (probe)
> return 0;
> @@ -3981,6 +3982,12 @@ static int reset_ivb_igd(struct pci_dev *dev, bool probe)
> if (!mmio_base)
> return -ENOMEM;
>
> + ret = pci_dev_reset_iommu_prepare(dev);
> + if (ret) {
> + pci_err(dev, "failed to stop IOMMU for a PCI reset: %d\n", ret);
> + goto out_iounmap;
> + }
> +
> iowrite32(0x00000002, mmio_base + MSG_CTL);
>
> /*
> @@ -4006,8 +4013,10 @@ static int reset_ivb_igd(struct pci_dev *dev, bool probe)
> reset_complete:
> iowrite32(0x00000002, mmio_base + NSDE_PWR_STATE);
>
> + pci_dev_reset_iommu_done(dev);
> +out_iounmap:
> pci_iounmap(dev, mmio_base);
> - return 0;
> + return ret;
> }
>
> /* Device-specific reset method for Chelsio T4-based adapters */
> @@ -4257,22 +4266,6 @@ static const struct pci_dev_reset_methods pci_dev_reset_methods[] = {
> { 0 }
> };
>
> -static int __pci_dev_specific_reset(struct pci_dev *dev, bool probe,
> - const struct pci_dev_reset_methods *i)
> -{
> - int ret;
> -
> - ret = pci_dev_reset_iommu_prepare(dev);
> - if (ret) {
> - pci_err(dev, "failed to stop IOMMU for a PCI reset: %d\n", ret);
> - return ret;
> - }
> -
> - ret = i->reset(dev, probe);
> - pci_dev_reset_iommu_done(dev);
> - return ret;
> -}
> -
> /*
> * These device-specific reset methods are here rather than in a driver
> * because when a host assigns a device to a guest VM, the host may need
> @@ -4287,7 +4280,7 @@ int pci_dev_specific_reset(struct pci_dev *dev, bool probe)
> i->vendor == (u16)PCI_ANY_ID) &&
> (i->device == dev->device ||
> i->device == (u16)PCI_ANY_ID))
> - return __pci_dev_specific_reset(dev, probe, i);
> + return i->reset(dev, probe);
> }
>
> return -ENOTTY;
> --
> 2.34.1
>
next prev parent reply other threads:[~2026-03-18 22:57 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-03-18 22:00 [PATCH rc] PCI: Fix nested pci_dev_reset_iommu_prepare/done() Nicolin Chen
2026-03-18 22:57 ` Bjorn Helgaas [this message]
2026-03-18 23:35 ` Nicolin Chen
2026-03-19 2:22 ` Tian, Kevin
2026-03-19 2:42 ` Nicolin Chen
2026-03-19 3:00 ` Tian, Kevin
2026-03-19 3:06 ` Nicolin Chen
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=20260318225705.GA479639@bhelgaas \
--to=helgaas@kernel.org \
--cc=bhelgaas@google.com \
--cc=jgg@nvidia.com \
--cc=joerg.roedel@amd.com \
--cc=kevin.tian@intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pci@vger.kernel.org \
--cc=nicolinc@nvidia.com \
--cc=xueshuai@linux.alibaba.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