Re: [PATCH v2] wifi: mt76: mt7921: Add PCIe AER handler support to prevent system crash

["Eason Lai (賴易聖)" <Eason.Lai@mediatek.com>]

On Wed, 2025-10-08 at 11:12 +0200, Johannes Berg wrote:
> Hi,
> 
> So ... yeah it's your driver, and yes it's already at v2, but ...
> this
> still seems very much papering over a problem?
> 
> On Wed, 2025-10-08 at 16:59 +0800, Mingyen Hsieh wrote:
> > 
> > Unable to handle kernel paging request at virtual address
> > ffffffc01099eac0
> > pc : mt76_dma_add_buf+0x124/0x188 [mt76]
> 
> Why does it even crash there?
> 

It's an out-of-range access: the HW returned an unexpected value that
wasn't validated before use.

v3 posted here:
https://lore.kernel.org/linux-wireless/20260506070458.3096180-1-jb.tsai@mediatek.com/

Thanks,
Eason

> > +++ b/drivers/net/wireless/mediatek/mt76/agg-rx.c
> > @@ -96,6 +96,9 @@ mt76_rx_aggr_reorder_work(struct work_struct
> > *work)
> >  	struct sk_buff_head frames;
> >  	int nframes;
> >  
> > +	if (atomic_read(&dev->bus_hung) == 1)
> > +		return;
> 
> And how does sprinkling this "magic dust" all over even do anything?
> 
> 
> This sounds a bit like the driver isn't able to deal with surprise
> removal, what happens if you e.g. rmmod it while running traffic? I
> believe the effect of the error recovery would be similar, if you
> don't
> handle it, so are you sure there's not a completely different
> underlying
> bug?
> 
> johannes