[PATCH char-misc-next] mei: bail out if no matching hbm request found

[PATCH char-misc-next] mei: bail out if no matching hbm request found

[Alexander Usyskin]

Umatched HBM client response acts on wrong client
as last traversed client is used.

Fill cl pointer only when match is found and add debug print
for that edge case.

Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
---
 drivers/misc/mei/hbm.c | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/drivers/misc/mei/hbm.c b/drivers/misc/mei/hbm.c
index a864ac7121b2..c93f3db34353 100644
--- a/drivers/misc/mei/hbm.c
+++ b/drivers/misc/mei/hbm.c
@@ -998,25 +998,26 @@ static void mei_hbm_cl_res(struct mei_device *dev,
 			   struct mei_hbm_cl_cmd *rs,
 			   enum mei_cb_file_ops fop_type)
 {
-	struct mei_cl *cl;
+	struct mei_cl *cl = NULL;
 	struct mei_cl_cb *cb, *next;
 
-	cl = NULL;
 	list_for_each_entry_safe(cb, next, &dev->ctrl_rd_list, list) {
 
-		cl = cb->cl;
-
 		if (cb->fop_type != fop_type)
 			continue;
 
-		if (mei_hbm_cl_addr_equal(cl, rs)) {
+		if (mei_hbm_cl_addr_equal(cb->cl, rs)) {
 			list_del_init(&cb->list);
+			cl = cb->cl;
 			break;
 		}
 	}
 
-	if (!cl)
+	if (!cl) {
+		dev_dbg(&dev->dev, "Can't find request for cmd %02d (%02d:%02d)\n",
+			rs->hbm_cmd, rs->host_addr, rs->me_addr);
 		return;
+	}
 
 	switch (fop_type) {
 	case MEI_FOP_CONNECT:

---
base-commit: 79e5797d243a466138f641932f81dc69f796060c
change-id: 20261101-bail_out_on_nonexisting_hbm-a9a578675835

Best regards,
-- 
Alexander Usyskin <alexander.usyskin@intel.com>

Re: [PATCH char-misc-next] mei: bail out if no matching hbm request found

[Greg Kroah-Hartman]

On Sun, Nov 01, 2026 at 04:32:06PM +0200, Alexander Usyskin wrote:
> Umatched HBM client response acts on wrong client
> as last traversed client is used.

Odd line wrapping :)

> Fill cl pointer only when match is found and add debug print
> for that edge case.
> 
> Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>

Does this need a Fixes and cc: stable tag, as it fixes a real issue
seen, or is it just a "handle the error path if it ever happens to be
hit" type of change for 7.5-rc1?

thanks,

greg k-h

RE: [PATCH char-misc-next] mei: bail out if no matching hbm request found

[Usyskin, Alexander]

> Subject: Re: [PATCH char-misc-next] mei: bail out if no matching hbm request
> found
> 
> On Sun, Nov 01, 2026 at 04:32:06PM +0200, Alexander Usyskin wrote:
> > Umatched HBM client response acts on wrong client
> > as last traversed client is used.
> 
> Odd line wrapping :)
> 
> > Fill cl pointer only when match is found and add debug print
> > for that edge case.
> >
> > Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
> 
> Does this need a Fixes and cc: stable tag, as it fixes a real issue
> seen, or is it just a "handle the error path if it ever happens to be
> hit" type of change for 7.5-rc1?
> 

This is "handle the error path" case.
Such failure should happen from time to time, but I never traced real issue to this.
I can't find exact commit that it fixes as it traces to the one of the first versions of the file.

- - 
Thanks,
Sasha


> thanks,
> 
> greg k-h