Re: [PATCH v2] PCI: Fix crash during pci_dev hot-unplug on pseries KVM guest

From: Lizhi Hou
Date: Mon Jul 29 2024 - 12:49:50 EST


Hi Amit

On 7/29/24 04:13, Amit Machhiwal wrote:
Hi Lizhi,

On 2024/07/26 11:45 AM, Lizhi Hou wrote:
On 7/26/24 10:52, Rob Herring wrote:
On Thu, Jul 25, 2024 at 6:06 PM Lizhi Hou <lizhi.hou@xxxxxxx> wrote:
Hi Amit,


I try to follow the option which add a OF flag. If Rob is ok with this,
I would suggest to use it instead of V1 patch

diff --git a/drivers/of/dynamic.c b/drivers/of/dynamic.c
index dda6092e6d3a..a401ed0463d9 100644
--- a/drivers/of/dynamic.c
+++ b/drivers/of/dynamic.c
@@ -382,6 +382,11 @@ void of_node_release(struct kobject *kobj)
__func__, node);
}

+ if (of_node_check_flag(node, OF_CREATED_WITH_CSET)) {
+ of_changeset_revert(node->data);
+ of_changeset_destroy(node->data);
+ }
What happens if multiple nodes are created in the changeset?
Ok. multiple nodes will not work.
+
if (node->child)
pr_err("ERROR: %s() unexpected children for %pOF/%s\n",
__func__, node->parent, node->full_name);
@@ -507,6 +512,7 @@ struct device_node *of_changeset_create_node(struct
of_changeset *ocs,
np = __of_node_dup(NULL, full_name);
if (!np)
return NULL;
+ of_node_set_flag(np, OF_CREATED_WITH_CSET);
This should be set where the data ptr is set.
Ok. It sounds the fix could be simplified to 3 lines change.
Thanks for the patch. The hot-plug and hot-unplug of PCI device seem to work
fine as expected. I see this patch would attempt to remove only the nodes which
were created in `of_pci_make_dev_node()` with the help of the newly introduced
flag, which looks good to me.

Also, since a call to `of_pci_make_dev_node()` from `pci_bus_add_device()`, that
creates devices nodes only for bridge devices, is conditional on
`pci_is_bridge()`, it only makes sense to retain the logical symmetry and call
`of_pci_remove_node()` conditionally on `pci_is_bridge()` as well in
`pci_stop_dev()`. Hence, I would like to propose the below change along with the
above patch:

diff --git a/drivers/pci/remove.c b/drivers/pci/remove.c
index 910387e5bdbf..c6394bf562cd 100644
--- a/drivers/pci/remove.c
+++ b/drivers/pci/remove.c
@@ -23,7 +23,8 @@ static void pci_stop_dev(struct pci_dev *dev)
device_release_driver(&dev->dev);
pci_proc_detach_device(dev);
pci_remove_sysfs_dev_files(dev);
- of_pci_remove_node(dev);
+ if (pci_is_bridge(dev))
+ of_pci_remove_node(dev);
pci_dev_assign_added(dev, false);
}

Please let me know of your thoughts on this and based on that I can spin the v3
of this patch.

As I mentioned, there are endpoints in pci quirks (pci/quirks.c) will also create nodes by of_pci_make_dev_node(). So please remove above two lines.


Thanks,

Lizhi


In addition to this, can this patch be taken as part of 6.11 as a bug fix?

Thanks,
Amit


diff --git a/drivers/pci/of.c b/drivers/pci/of.c
index 51e3dd0ea5ab..0b3ba1e1b18c 100644
--- a/drivers/pci/of.c
+++ b/drivers/pci/of.c
@@ -613,7 +613,7 @@ void of_pci_remove_node(struct pci_dev *pdev)
        struct device_node *np;

        np = pci_device_to_OF_node(pdev);
-       if (!np || !of_node_check_flag(np, OF_DYNAMIC))
+       if (!np || !of_node_check_flag(np, OF_CREATED_WITH_CSET))
                return;
        pdev->dev.of_node = NULL;

@@ -672,6 +672,7 @@ void of_pci_make_dev_node(struct pci_dev *pdev)
        if (ret)
                goto out_free_node;

+       of_node_set_flag(np, OF_CREATED_WITH_CSET);
        np->data = cset;
        pdev->dev.of_node = np;
        kfree(name);
diff --git a/include/linux/of.h b/include/linux/of.h
index a0bedd038a05..a46317f6626e 100644
--- a/include/linux/of.h
+++ b/include/linux/of.h
@@ -153,6 +153,7 @@ extern struct device_node *of_stdout;
 #define OF_POPULATED_BUS       4 /* platform bus created for children */
 #define OF_OVERLAY             5 /* allocated for an overlay */
 #define OF_OVERLAY_FREE_CSET   6 /* in overlay cset being freed */
+#define OF_CREATED_WITH_CSET    7 /* created by of_changeset_create_node */

 #define OF_BAD_ADDR    ((u64)-1)


Lizhi

Rob