Re: [PATCH] NVMe: Don't unmap controller registers on reset

[Gabriel Krisman Bertazi <krisman@linux.vnet.ibm.com>]

Jiri Slaby <jslaby@suse.cz> writes:

>
>> @@ -2734,18 +2724,21 @@ static int nvme_dev_map(struct nvme_dev *dev)
>>  
>>  	return 0;
>>  
>> - unmap:
>> -	iounmap(dev->bar);
>> -	dev->bar = NULL;
>>   disable:
>>  	pci_release_regions(pdev);
>> - disable_pci:
>> -	pci_disable_device(pdev);
>
> Is this a correct backport?
>
> The original removes pci_release_regions, not pci_disable_device.
>

Hi Jiri,

Oh, you are correct.  We need pci_disable_device here to reverse
pci_enable_device_mem on error, while should go pci_release_regions into
nvme_dev_map to reverse the pci_request_selected_regions call.

Unfortunately, my test case never touched this error path.

Greg, would you take the fix below into the -stable tree?

-- >8 --
Subject: [PATCH] nvme: Call pci_disable_device on the error path.

Commit 5706aca74fe4 ("NVMe: Don't unmap controller registers on reset"),
which backported b00a726a9fd8 to the 4.4.y kernel introduced a
regression in which it didn't call pci_disable_device in the error path
of nvme_pci_enable.

Reported-by: Jiri Slaby <jslaby@suse.cz>
Embarassed-developer: Gabriel Krisman Bertazi <krisman@linux.vnet.ibm.com>
Signed-off-by: Gabriel Krisman Bertazi <krisman@linux.vnet.ibm.com>
---
 drivers/nvme/host/pci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index 289a5df0d44a..c851bc53831c 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -2725,7 +2725,7 @@ static int nvme_pci_enable(struct nvme_dev *dev)
 	return 0;
 
  disable:
-	pci_release_regions(pdev);
+	pci_disable_device(pdev);
 
 	return result;
 }
-- 
2.7.4