[PATCH] infiniband: change flags from int to long

[PATCH] infiniband: change flags from int to long

[Steven Rostedt]

It is a bug to have irq saved flags as an int and not long since
some archs may use more that 32 bits in flags.

(This patch was only compiled tested)

[
 Found by the -rt patch checks. These should now be in mainline,
 but it looks like they may not have been used.
]

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
---
 drivers/infiniband/hw/ipath/ipath_verbs.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

Index: linus.git/drivers/infiniband/hw/ipath/ipath_verbs.c
===================================================================
--- linus.git.orig/drivers/infiniband/hw/ipath/ipath_verbs.c	2008-08-12 10:23:25.000000000 -0400
+++ linus.git/drivers/infiniband/hw/ipath/ipath_verbs.c	2008-08-12 10:25:24.000000000 -0400
@@ -1021,7 +1021,7 @@ static void sdma_complete(void *cookie, 
 	struct ipath_verbs_txreq *tx = cookie;
 	struct ipath_qp *qp = tx->qp;
 	struct ipath_ibdev *dev = to_idev(qp->ibqp.device);
-	unsigned int flags;
+	unsigned long flags;
 	enum ib_wc_status ibs = status == IPATH_SDMA_TXREQ_S_OK ?
 		IB_WC_SUCCESS : IB_WC_WR_FLUSH_ERR;
 
@@ -1051,7 +1051,7 @@ static void sdma_complete(void *cookie, 
 
 static void decrement_dma_busy(struct ipath_qp *qp)
 {
-	unsigned int flags;
+	unsigned long flags;
 
 	if (atomic_dec_and_test(&qp->s_dma_busy)) {
 		spin_lock_irqsave(&qp->s_lock, flags);
@@ -1221,7 +1221,7 @@ static int ipath_verbs_send_pio(struct i
 	unsigned flush_wc;
 	u32 control;
 	int ret;
-	unsigned int flags;
+	unsigned long flags;
 
 	piobuf = ipath_getpiobuf(dd, plen, NULL);
 	if (unlikely(piobuf == NULL)) {

Re: [PATCH] infiniband: change flags from int to long

[Roland Dreier]

 > It is a bug to have irq saved flags as an int and not long since
 > some archs may use more that 32 bits in flags.

Isn't this already upstream for a while as 52fd8ca6?

 - R.

Re: [PATCH] infiniband: change flags from int to long

[Steven Rostedt]

On Tue, 12 Aug 2008, Roland Dreier wrote:

>  > It is a bug to have irq saved flags as an int and not long since
>  > some archs may use more that 32 bits in flags.
> 
> Isn't this already upstream for a while as 52fd8ca6?

This is the problem with having multiple git repos lying around. You never 
know which one is updated. I should have done a git pull on the git repo I 
examined.

Last commit on the repo I looked at:

commit 94ad374a0751f40d25e22e036c37f7263569d24c
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date:   Wed Jul 30 14:45:12 2008 -0700


And the commit you stated:

commit 52fd8ca6ad4124c15952ded35cfcf6adbd7ae8d4
Author: Vegard Nossum <vegard.nossum@gmail.com>
Date:   Wed Jul 30 09:29:06 2008 -0700


Which Linus pulled in on:

commit 273b2578392bbf6e5c47a8a3d1ee461ce6fc7182
Merge: 8e43e12... 06a91a0...
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date:   Thu Aug 7 18:14:07 2008 -0700


Sorry for the noise :-/

-- Steve