diff mbox series

[RFC,v1,1/2] block: bsg: resume scsi device before accessing

Message ID c04a11a590628c2497cef113b0dfea781de90416.1611719814.git.asutoshd@codeaurora.org (mailing list archive)
State New
Headers show
Series Fix deadlock in ufs | expand

Commit Message

Asutosh Das Jan. 27, 2021, 4 a.m. UTC
Resumes the scsi device before accessing it.

Change-Id: I2929af60f2a92c89704a582fcdb285d35b429fde
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
Signed-off-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Bao D. Nguyen <nguyenb@codeaurora.org>
---
 block/bsg.c | 8 ++++++++
 1 file changed, 8 insertions(+)

Comments

Avri Altman Jan. 27, 2021, 7:09 a.m. UTC | #1
> 
> Resumes the scsi device before accessing it.
> 
> Change-Id: I2929af60f2a92c89704a582fcdb285d35b429fde
> Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
> Signed-off-by: Can Guo <cang@codeaurora.org>
> Signed-off-by: Bao D. Nguyen <nguyenb@codeaurora.org>
Following this patch, is it possible to revert commit 74e5e468b664d?

Thanks,
Avri
Can Guo Jan. 27, 2021, 7:59 a.m. UTC | #2
On 2021-01-27 15:09, Avri Altman wrote:
>> 
>> Resumes the scsi device before accessing it.
>> 
>> Change-Id: I2929af60f2a92c89704a582fcdb285d35b429fde
>> Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
>> Signed-off-by: Can Guo <cang@codeaurora.org>
>> Signed-off-by: Bao D. Nguyen <nguyenb@codeaurora.org>
> Following this patch, is it possible to revert commit 74e5e468b664d?
> 

No, but this is a good finding... This change assumes
that the queue->queue_data is a scsi_device, which is
why we call scsi_auto_pm_get(). But for ufs_bsg's case,
queue->queue_data is a device...

Thanks,
Can Guo.

> Thanks,
> Avri
Can Guo Jan. 27, 2021, 8:53 a.m. UTC | #3
On 2021-01-27 15:59, Can Guo wrote:
> On 2021-01-27 15:09, Avri Altman wrote:
>>> 
>>> Resumes the scsi device before accessing it.
>>> 
>>> Change-Id: I2929af60f2a92c89704a582fcdb285d35b429fde
>>> Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
>>> Signed-off-by: Can Guo <cang@codeaurora.org>
>>> Signed-off-by: Bao D. Nguyen <nguyenb@codeaurora.org>
>> Following this patch, is it possible to revert commit 74e5e468b664d?
>> 
> 
> No, but this is a good finding... This change assumes
> that the queue->queue_data is a scsi_device, which is
> why we call scsi_auto_pm_get(). But for ufs_bsg's case,
> queue->queue_data is a device...
> 

If we call pm_runtime_get/put_sync(bcd->class_dev->parent) in
bsg_get/put_device(), commit 74e5e468b664d can be reverted.
This is just a rough idea.

> Thanks,
> Can Guo.
> 
>> Thanks,
>> Avri
Bart Van Assche Feb. 7, 2021, 2:23 a.m. UTC | #4
On 1/26/21 8:00 PM, Asutosh Das wrote:
> Resumes the scsi device before accessing it.
> 
> Change-Id: I2929af60f2a92c89704a582fcdb285d35b429fde
> Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
> Signed-off-by: Can Guo <cang@codeaurora.org>
> Signed-off-by: Bao D. Nguyen <nguyenb@codeaurora.org>

No Change-Id tags in upstream patches please.

Thanks,

Bart.
diff mbox series

Patch

diff --git a/block/bsg.c b/block/bsg.c
index d7bae94..f4c197f 100644
--- a/block/bsg.c
+++ b/block/bsg.c
@@ -306,12 +306,16 @@  static struct bsg_device *bsg_get_device(struct inode *inode, struct file *file)
 static int bsg_open(struct inode *inode, struct file *file)
 {
 	struct bsg_device *bd;
+	struct scsi_device *sd;
 
 	bd = bsg_get_device(inode, file);
 
 	if (IS_ERR(bd))
 		return PTR_ERR(bd);
 
+	sd = (struct scsi_device *) bd->queue->queuedata;
+	if (scsi_autopm_get_device(sd))
+		return -EIO;
 	file->private_data = bd;
 	return 0;
 }
@@ -319,8 +323,12 @@  static int bsg_open(struct inode *inode, struct file *file)
 static int bsg_release(struct inode *inode, struct file *file)
 {
 	struct bsg_device *bd = file->private_data;
+	struct scsi_device *sd;
 
 	file->private_data = NULL;
+	sd = (struct scsi_device *) bd->queue->queuedata;
+	scsi_autopm_put_device(sd);
+
 	return bsg_put_device(bd);
 }