When MSI-X initialization fails, vPCI hides the capability, but
removing handlers and datas won't be performed until the device is
deassigned. So, implement MSI-X cleanup hook that will be called
to cleanup MSI-X related handlers and free it's associated data when
initialization fails.

Since cleanup function of MSI-X is implemented, delete the open-code
in vpci_deassign_device().

Signed-off-by: Jiqian Chen <Jiqian.Chen@amd.com>
---
cc: "Roger Pau Monné" <roger.pau@citrix.com>
---
v11->v12 changes:
* In cleanup_msix(), move check "if ( !hide )" above vpci_remove_registers().
* Remove the check "!pdev->msix_pos" since current callers already do that.

v10->v11 changes:
* Move calling all cleanup hook in vpci_deassign_device() out of this patch.
* Add hide parameter to cleanup_msix().
* Check hide, if it is false, return directly instead of letting ctrl RO.

v9->v10 changes:
* Call all cleanup hook in vpci_deassign_device() instead of cleanup_msix().

v8->v9 changes:
* Modify commit message.
* Call cleanup_msix() in vpci_deassign_device() to remove the open-code to cleanup msix datas.
* In cleanup_msix(), move "list_del(&vpci->msix->next);" above for loop of iounmap msix tables.

v7->v8 changes:
* Given the code in vpci_remove_registers() an error in the removal of
  registers would likely imply memory corruption, at which point it's
  best to fully disable the device. So, Rollback the last two modifications of v7.

v6->v7 changes:
* Change the pointer parameter of cleanup_msix() to be const.
* When vpci_remove_registers() in cleanup_msix() fails, not to return
  directly, instead try to free msix and re-add ctrl handler.
* Pass pdev->vpci into vpci_add_register() instead of pdev->vpci->msix in
  init_msix() since we need that every handler realize that msix is NULL
  when msix is freed but handlers are still in there.

v5->v6 changes:
* Change the logic to add dummy handler when !vpci->msix in cleanup_msix().

v4->v5 changes:
* Change definition "static void cleanup_msix" to "static int cf_check cleanup_msix"
  since cleanup hook is changed to be int.
* Add a read-only register for MSIX Control Register in the end of cleanup_msix().

v3->v4 changes:
* Change function name from fini_msix() to cleanup_msix().
* Change to use XFREE to free vpci->msix.
* In cleanup function, change the sequence of check and remove action according to
  init_msix().

v2->v3 changes:
* Remove unnecessary clean operations in fini_msix().

v1->v2 changes:
new patch.

Best regards,
Jiqian Chen.
---
 xen/drivers/vpci/msix.c | 44 ++++++++++++++++++++++++++++++++++++++++-
 xen/drivers/vpci/vpci.c |  8 --------
 2 files changed, 43 insertions(+), 9 deletions(-)

diff --git a/xen/drivers/vpci/msix.c b/xen/drivers/vpci/msix.c
index 032e471bb1c0..8dcf2cf9d598 100644
--- a/xen/drivers/vpci/msix.c
+++ b/xen/drivers/vpci/msix.c
@@ -656,6 +656,48 @@ int vpci_make_msix_hole(const struct pci_dev *pdev)
     return 0;
 }
 
+static int cf_check cleanup_msix(const struct pci_dev *pdev, bool hide)
+{
+    int rc;
+    struct vpci *vpci = pdev->vpci;
+    const unsigned int msix_pos = pdev->msix_pos;
+
+    if ( vpci->msix )
+    {
+        list_del(&vpci->msix->next);
+        for ( unsigned int i = 0; i < ARRAY_SIZE(vpci->msix->table); i++ )
+            if ( vpci->msix->table[i] )
+                iounmap(vpci->msix->table[i]);
+
+        XFREE(vpci->msix);
+    }
+
+    if ( !hide )
+        return 0;
+
+    rc = vpci_remove_registers(vpci, msix_control_reg(msix_pos), 2);
+    if ( rc )
+    {
+        printk(XENLOG_ERR "%pd %pp: fail to remove MSIX handlers rc=%d\n",
+               pdev->domain, &pdev->sbdf, rc);
+        ASSERT_UNREACHABLE();
+        return rc;
+    }
+
+    /*
+     * The driver may not traverse the capability list and think device
+     * supports MSIX by default. So here let the control register of MSIX
+     * be Read-Only is to ensure MSIX disabled.
+     */
+    rc = vpci_add_register(vpci, vpci_hw_read16, NULL,
+                           msix_control_reg(msix_pos), 2, NULL);
+    if ( rc )
+        printk(XENLOG_ERR "%pd %pp: fail to add MSIX ctrl handler rc=%d\n",
+               pdev->domain, &pdev->sbdf, rc);
+
+    return rc;
+}
+
 static int cf_check init_msix(struct pci_dev *pdev)
 {
     struct domain *d = pdev->domain;
@@ -751,7 +793,7 @@ static int cf_check init_msix(struct pci_dev *pdev)
      */
     return vpci_make_msix_hole(pdev);
 }
-REGISTER_VPCI_CAP(MSIX, init_msix, NULL);
+REGISTER_VPCI_CAP(MSIX, init_msix, cleanup_msix);
 
 /*
  * Local variables:
diff --git a/xen/drivers/vpci/vpci.c b/xen/drivers/vpci/vpci.c
index 7aaf015f63d4..3c9bebcbe977 100644
--- a/xen/drivers/vpci/vpci.c
+++ b/xen/drivers/vpci/vpci.c
@@ -356,18 +356,10 @@ void vpci_deassign_device(struct pci_dev *pdev)
         xfree(r);
     }
     spin_unlock(&pdev->vpci->lock);
-    if ( pdev->vpci->msix )
-    {
-        list_del(&pdev->vpci->msix->next);
-        for ( i = 0; i < ARRAY_SIZE(pdev->vpci->msix->table); i++ )
-            if ( pdev->vpci->msix->table[i] )
-                iounmap(pdev->vpci->msix->table[i]);
-    }
 
     for ( i = 0; i < ARRAY_SIZE(pdev->vpci->header.bars); i++ )
         rangeset_destroy(pdev->vpci->header.bars[i].mem);
 
-    xfree(pdev->vpci->msix);
     xfree(pdev->vpci);
     pdev->vpci = NULL;
 }
-- 
2.34.1

On Mon, Dec 08, 2025 at 04:18:15PM +0800, Jiqian Chen wrote:
> When MSI-X initialization fails, vPCI hides the capability, but
> removing handlers and datas won't be performed until the device is
> deassigned. So, implement MSI-X cleanup hook that will be called
> to cleanup MSI-X related handlers and free it's associated data when
> initialization fails.
> 
> Since cleanup function of MSI-X is implemented, delete the open-code
> in vpci_deassign_device().
> 
> Signed-off-by: Jiqian Chen <Jiqian.Chen@amd.com>
> ---
> cc: "Roger Pau Monné" <roger.pau@citrix.com>
> ---
> v11->v12 changes:
> * In cleanup_msix(), move check "if ( !hide )" above vpci_remove_registers().
> * Remove the check "!pdev->msix_pos" since current callers already do that.
> 
> v10->v11 changes:
> * Move calling all cleanup hook in vpci_deassign_device() out of this patch.
> * Add hide parameter to cleanup_msix().
> * Check hide, if it is false, return directly instead of letting ctrl RO.
> 
> v9->v10 changes:
> * Call all cleanup hook in vpci_deassign_device() instead of cleanup_msix().
> 
> v8->v9 changes:
> * Modify commit message.
> * Call cleanup_msix() in vpci_deassign_device() to remove the open-code to cleanup msix datas.
> * In cleanup_msix(), move "list_del(&vpci->msix->next);" above for loop of iounmap msix tables.
> 
> v7->v8 changes:
> * Given the code in vpci_remove_registers() an error in the removal of
>   registers would likely imply memory corruption, at which point it's
>   best to fully disable the device. So, Rollback the last two modifications of v7.
> 
> v6->v7 changes:
> * Change the pointer parameter of cleanup_msix() to be const.
> * When vpci_remove_registers() in cleanup_msix() fails, not to return
>   directly, instead try to free msix and re-add ctrl handler.
> * Pass pdev->vpci into vpci_add_register() instead of pdev->vpci->msix in
>   init_msix() since we need that every handler realize that msix is NULL
>   when msix is freed but handlers are still in there.
> 
> v5->v6 changes:
> * Change the logic to add dummy handler when !vpci->msix in cleanup_msix().
> 
> v4->v5 changes:
> * Change definition "static void cleanup_msix" to "static int cf_check cleanup_msix"
>   since cleanup hook is changed to be int.
> * Add a read-only register for MSIX Control Register in the end of cleanup_msix().
> 
> v3->v4 changes:
> * Change function name from fini_msix() to cleanup_msix().
> * Change to use XFREE to free vpci->msix.
> * In cleanup function, change the sequence of check and remove action according to
>   init_msix().
> 
> v2->v3 changes:
> * Remove unnecessary clean operations in fini_msix().
> 
> v1->v2 changes:
> new patch.
> 
> Best regards,
> Jiqian Chen.
> ---
>  xen/drivers/vpci/msix.c | 44 ++++++++++++++++++++++++++++++++++++++++-
>  xen/drivers/vpci/vpci.c |  8 --------
>  2 files changed, 43 insertions(+), 9 deletions(-)
> 
> diff --git a/xen/drivers/vpci/msix.c b/xen/drivers/vpci/msix.c
> index 032e471bb1c0..8dcf2cf9d598 100644
> --- a/xen/drivers/vpci/msix.c
> +++ b/xen/drivers/vpci/msix.c
> @@ -656,6 +656,48 @@ int vpci_make_msix_hole(const struct pci_dev *pdev)
>      return 0;
>  }
>  
> +static int cf_check cleanup_msix(const struct pci_dev *pdev, bool hide)
> +{
> +    int rc;
> +    struct vpci *vpci = pdev->vpci;
> +    const unsigned int msix_pos = pdev->msix_pos;
> +
> +    if ( vpci->msix )
> +    {
> +        list_del(&vpci->msix->next);
> +        for ( unsigned int i = 0; i < ARRAY_SIZE(vpci->msix->table); i++ )
> +            if ( vpci->msix->table[i] )
> +                iounmap(vpci->msix->table[i]);
> +
> +        XFREE(vpci->msix);
> +    }
> +
> +    if ( !hide )
> +        return 0;
> +
> +    rc = vpci_remove_registers(vpci, msix_control_reg(msix_pos), 2);
> +    if ( rc )
> +    {
> +        printk(XENLOG_ERR "%pd %pp: fail to remove MSIX handlers rc=%d\n",
> +               pdev->domain, &pdev->sbdf, rc);
> +        ASSERT_UNREACHABLE();
> +        return rc;
> +    }
> +
> +    /*
> +     * The driver may not traverse the capability list and think device
> +     * supports MSIX by default. So here let the control register of MSIX
> +     * be Read-Only is to ensure MSIX disabled.
> +     */
> +    rc = vpci_add_register(vpci, vpci_hw_read16, NULL,
> +                           msix_control_reg(msix_pos), 2, NULL);
> +    if ( rc )
> +        printk(XENLOG_ERR "%pd %pp: fail to add MSIX ctrl handler rc=%d\n",
> +               pdev->domain, &pdev->sbdf, rc);

Like the previous patch, I don't think this last bit is relevant for
domUs?  Only the hardware domain needs to have the control register
explicitly handled.

I don't mind adjusting at commit if we agree.

Reviewed-by: Roger Pau Monné <roger.pau@citrix.com>

Thanks, Roger.

On 2026/1/21 17:25, Roger Pau Monné wrote:
> On Mon, Dec 08, 2025 at 04:18:15PM +0800, Jiqian Chen wrote:
>> When MSI-X initialization fails, vPCI hides the capability, but
>> removing handlers and datas won't be performed until the device is
>> deassigned. So, implement MSI-X cleanup hook that will be called
>> to cleanup MSI-X related handlers and free it's associated data when
>> initialization fails.
>>
>> Since cleanup function of MSI-X is implemented, delete the open-code
>> in vpci_deassign_device().
>>
>> Signed-off-by: Jiqian Chen <Jiqian.Chen@amd.com>
>> ---
>> cc: "Roger Pau Monné" <roger.pau@citrix.com>
>> ---
>> v11->v12 changes:
>> * In cleanup_msix(), move check "if ( !hide )" above vpci_remove_registers().
>> * Remove the check "!pdev->msix_pos" since current callers already do that.
>>
>> v10->v11 changes:
>> * Move calling all cleanup hook in vpci_deassign_device() out of this patch.
>> * Add hide parameter to cleanup_msix().
>> * Check hide, if it is false, return directly instead of letting ctrl RO.
>>
>> v9->v10 changes:
>> * Call all cleanup hook in vpci_deassign_device() instead of cleanup_msix().
>>
>> v8->v9 changes:
>> * Modify commit message.
>> * Call cleanup_msix() in vpci_deassign_device() to remove the open-code to cleanup msix datas.
>> * In cleanup_msix(), move "list_del(&vpci->msix->next);" above for loop of iounmap msix tables.
>>
>> v7->v8 changes:
>> * Given the code in vpci_remove_registers() an error in the removal of
>>   registers would likely imply memory corruption, at which point it's
>>   best to fully disable the device. So, Rollback the last two modifications of v7.
>>
>> v6->v7 changes:
>> * Change the pointer parameter of cleanup_msix() to be const.
>> * When vpci_remove_registers() in cleanup_msix() fails, not to return
>>   directly, instead try to free msix and re-add ctrl handler.
>> * Pass pdev->vpci into vpci_add_register() instead of pdev->vpci->msix in
>>   init_msix() since we need that every handler realize that msix is NULL
>>   when msix is freed but handlers are still in there.
>>
>> v5->v6 changes:
>> * Change the logic to add dummy handler when !vpci->msix in cleanup_msix().
>>
>> v4->v5 changes:
>> * Change definition "static void cleanup_msix" to "static int cf_check cleanup_msix"
>>   since cleanup hook is changed to be int.
>> * Add a read-only register for MSIX Control Register in the end of cleanup_msix().
>>
>> v3->v4 changes:
>> * Change function name from fini_msix() to cleanup_msix().
>> * Change to use XFREE to free vpci->msix.
>> * In cleanup function, change the sequence of check and remove action according to
>>   init_msix().
>>
>> v2->v3 changes:
>> * Remove unnecessary clean operations in fini_msix().
>>
>> v1->v2 changes:
>> new patch.
>>
>> Best regards,
>> Jiqian Chen.
>> ---
>>  xen/drivers/vpci/msix.c | 44 ++++++++++++++++++++++++++++++++++++++++-
>>  xen/drivers/vpci/vpci.c |  8 --------
>>  2 files changed, 43 insertions(+), 9 deletions(-)
>>
>> diff --git a/xen/drivers/vpci/msix.c b/xen/drivers/vpci/msix.c
>> index 032e471bb1c0..8dcf2cf9d598 100644
>> --- a/xen/drivers/vpci/msix.c
>> +++ b/xen/drivers/vpci/msix.c
>> @@ -656,6 +656,48 @@ int vpci_make_msix_hole(const struct pci_dev *pdev)
>>      return 0;
>>  }
>>  
>> +static int cf_check cleanup_msix(const struct pci_dev *pdev, bool hide)
>> +{
>> +    int rc;
>> +    struct vpci *vpci = pdev->vpci;
>> +    const unsigned int msix_pos = pdev->msix_pos;
>> +
>> +    if ( vpci->msix )
>> +    {
>> +        list_del(&vpci->msix->next);
>> +        for ( unsigned int i = 0; i < ARRAY_SIZE(vpci->msix->table); i++ )
>> +            if ( vpci->msix->table[i] )
>> +                iounmap(vpci->msix->table[i]);
>> +
>> +        XFREE(vpci->msix);
>> +    }
>> +
>> +    if ( !hide )
>> +        return 0;
>> +
>> +    rc = vpci_remove_registers(vpci, msix_control_reg(msix_pos), 2);
>> +    if ( rc )
>> +    {
>> +        printk(XENLOG_ERR "%pd %pp: fail to remove MSIX handlers rc=%d\n",
>> +               pdev->domain, &pdev->sbdf, rc);
>> +        ASSERT_UNREACHABLE();
>> +        return rc;
>> +    }
>> +
>> +    /*
>> +     * The driver may not traverse the capability list and think device
>> +     * supports MSIX by default. So here let the control register of MSIX
>> +     * be Read-Only is to ensure MSIX disabled.
>> +     */
>> +    rc = vpci_add_register(vpci, vpci_hw_read16, NULL,
>> +                           msix_control_reg(msix_pos), 2, NULL);
>> +    if ( rc )
>> +        printk(XENLOG_ERR "%pd %pp: fail to add MSIX ctrl handler rc=%d\n",
>> +               pdev->domain, &pdev->sbdf, rc);
> 
> Like the previous patch, I don't think this last bit is relevant for
> domUs?  Only the hardware domain needs to have the control register
> explicitly handled.
> 
> I don't mind adjusting at commit if we agree.
I agree with you.
Thank you for help to make changes of this and previous patch when you submit.

> 
> Reviewed-by: Roger Pau Monné <roger.pau@citrix.com>
> 
> Thanks, Roger.

-- 
Best regards,
Jiqian Chen.