Commit 07b61e0c authored Jul 11, 2016 by Abdulla Anam Committed by Gerrit - the friendly Code Review server Jul 12, 2016

msm: vidc: check for venus_hfi_device state in resume



Now since resume is called to ensure venus is in power on state
during scale clocks, an issue is seen when client responds to an
error notification with stream_off. This causes scale clocks to
be called which powers_on venus and enable regulators. But an
earlier sys_error deinited all resources and hence a null pointer
exception is seen causing kernel panic.

CRs-Fixed: 1040026
Change-Id: Ibb660397f8e7e3e0cd0a5e672924925e605b2eb7
Signed-off-by: Abdulla Anam <abdullahanam@codeaurora.org>

parent e99d73e0

drivers/media/platform/msm/vidc/venus_hfi.c

+3 −0

Original line number	Original line	Diff line number	Diff line
	@@ -4271,6 +4271,9 @@ static inline int __resume(struct venus_hfi_device *device)
	} else if (device->power_enabled) {		} else if (device->power_enabled) {
	dprintk(VIDC_DBG, "Power is already enabled\n");		dprintk(VIDC_DBG, "Power is already enabled\n");
	goto exit;		goto exit;
			} else if (!__core_in_valid_state(device)) {
			dprintk(VIDC_DBG, "venus_hfi_device in deinit state.");
			return -EINVAL;
	}		}

	dprintk(VIDC_DBG, "Resuming from power collapse\n");		dprintk(VIDC_DBG, "Resuming from power collapse\n");