This patch fixes the following two erroneous error paths:
hw_usecnt is allocated with a value of 0, therefore in an early error case,
calling sh_mobile_lcdc_clk_off() will wrongly conclude, that hw_usecnt has
already been incremented. Then sh_mobile_lcdc_runtime_suspend() will be called,
which will access uninitialised data fields and crash the kernel.
sh_mobile_lcdc_stop() can be called before framebuffer has been allocated, then
ch->info is NULL and dereferencing it will Oops too.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
* flush frame, and wait for frame end interrupt
* clean up deferred io and enable clock
*/
- if (ch->info->fbdefio) {
+ if (ch->info && ch->info->fbdefio) {
ch->frame_end = 0;
schedule_delayed_work(&ch->info->deferred_work, 0);
wait_event(ch->frame_end_wait, ch->frame_end);
return PTR_ERR(priv->dot_clk);
}
}
- atomic_set(&priv->hw_usecnt, -1);
/* Runtime PM support involves two step for this driver:
* 1) Enable Runtime PM
priv->irq = i;
pdata = pdev->dev.platform_data;
+ atomic_set(&priv->hw_usecnt, -1);
j = 0;
for (i = 0; i < ARRAY_SIZE(pdata->ch); i++) {