On Fri, 14 May 2010 10:30:47 -0400
James Smart <james.smart@xxxxxxxxxx> wrote:
> Actually, it may be just checking the file->private_data pointer for NULL at
> the file entry points.
bsg clears file->private_data in bsg_release() so nobody should not
see file->private_data NULL pointer...
> Although, the minor should not be deallocated from bsg
> until all releases are called.
bsg deallocate minor in bsg_unregister_queue(). It means, for example,
after a sas LLD frees a remote port, user-space application can't
access to it. I guess that it's the right thing.
The tricky part is, when a sas LLD frees a remote port, there might be
some user-space applications that still open a bsg device. So you
can't call blk_cleanup_queue() at that time.
You might hit the similar problem to the commit
93c20a59af4624aedf53f8320606b355aa951bc1. The following fix works?
=
From: FUJITA Tomonori <fujita.tomonori@xxxxxxxxxxxxx>
Subject: [PATCH] scsi_transport_fc: fix the lifetime of sas bsg objects
fc_bsg_remove can't call blk_cleanup_queue() since there might be
applications that open a fc_host (or rport).
The commit 93c20a59af4624aedf53f8320606b355aa951bc1 fixed the same
lifetime problem of scsi_transport_sas.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@xxxxxxxxxxxxx>
---
drivers/scsi/scsi_transport_fc.c | 17 +++++++++++++----
1 files changed, 13 insertions(+), 4 deletions(-)
diff --git a/drivers/scsi/scsi_transport_fc.c b/drivers/scsi/scsi_transport_fc.c
index 0681378..885b26b 100644
--- a/drivers/scsi/scsi_transport_fc.c
+++ b/drivers/scsi/scsi_transport_fc.c
@@ -1897,6 +1897,10 @@ static int fc_target_match(struct attribute_container *cont,
static void fc_rport_dev_release(struct device *dev)
{
struct fc_rport *rport = dev_to_rport(dev);
+
+ if (rport->rqst_q)
+ blk_cleanup_queue(rport->rqst_q);
+
put_device(dev->parent);
kfree(rport);
}
@@ -3944,6 +3948,13 @@ fc_bsg_rport_handler(struct request_queue *q)
fc_bsg_request_handler(q, shost, rport, &rport->dev);
}
+static void fc_host_bsg_release(struct device *dev)
+{
+ struct fc_host_attrs *fc_host = shost_to_fc_host(dev_to_shost(dev));
+
+ if (fc_host->rqst_q)
+ blk_cleanup_queue(fc_host->rqst_q);
+}
/**
* fc_bsg_hostadd - Create and add the bsg hooks so we can receive requests
@@ -3981,7 +3992,7 @@ fc_bsg_hostadd(struct Scsi_Host *shost, struct fc_host_attrs *fc_host)
blk_queue_rq_timed_out(q, fc_bsg_job_timeout);
blk_queue_rq_timeout(q, FC_DEFAULT_BSG_TIMEOUT);
- err = bsg_register_queue(q, dev, bsg_name, NULL);
+ err = bsg_register_queue(q, dev, bsg_name, fc_host_bsg_release);
if (err) {
printk(KERN_ERR "fc_host%d: bsg interface failed to "
"initialize - register queue\n",
@@ -4048,10 +4059,8 @@ fc_bsg_rportadd(struct Scsi_Host *shost, struct fc_rport *rport)
static void
fc_bsg_remove(struct request_queue *q)
{
- if (q) {
+ if (q)
bsg_unregister_queue(q);
- blk_cleanup_queue(q);
- }
}
--
1.6.5
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
