Skip to contentSkip to conversation

fpga: region: release bridges in successful path

PATCH In review Vyacheslav V. Yurkov 2 participants 2 messages Sep 27, 2017 FPGA manager framework

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fpga: region: release bridges in successful path

From: Vyacheslav Yurkov <[email protected]>

In case of successful FPGA reconfiguration bridges mutexes are not unlocked properly.

---
 drivers/fpga/fpga-region.c | 1 +
 1 file changed, 1 insertion(+)
drivers/fpga/fpga-region.c +1 -0
index d9ab7c7..ed858e3 100644
@@ -274,6 +274,7 @@ static int fpga_region_program_fpga(struct fpga_region *region,
 	}
 
 	fpga_mgr_put(mgr);
+	fpga_bridges_put(&region->bridge_list);
 	fpga_region_put(region);
 
 	return 0;
Signature · 1 line
2.7.4
#1 lore raw

On Wed, Sep 27, 2017 at 11:43 AM, Vyacheslav V. Yurkov <[email protected]> wrote:

Hi Vyacheslav,

Thanks for the patch, but actually this is on purpose to prevent someone disabling the bridges when a region has hardware running in it. The fpga_bridges_put happens when the device tree overlay is removed in fpga_region_notify_post_remove(). When an overlay programs a region, the region gets exclusive control of the bridges and disables them before programming. After programming, the region re-enables the bridges and keeps exclusive control of them. Some drivers are loaded by device tree and are expecting those bridges to stay enabled. Later when the overlay is removed, the drivers are unloaded first, the bridges are disable and put.

Alan

Quoted 25 lines

From: Vyacheslav Yurkov <[email protected]>

In case of successful FPGA reconfiguration bridges mutexes are not unlocked properly.

---
 drivers/fpga/fpga-region.c | 1 +
 1 file changed, 1 insertion(+)
drivers/fpga/fpga-region.c +1 -1
index d9ab7c7..ed858e3 100644
@@ -274,6 +274,7 @@ static int fpga_region_program_fpga(struct fpga_region *region,
        }
 
        fpga_mgr_put(mgr);
+       fpga_bridges_put(&region->bridge_list);
        fpga_region_put(region);
 
        return 0;
--

2.7.4