Re: [PATCH v2 2/2] wifi: ath11k: Implement sram dump interface

[Baochen Qiang <quic_bqiang@xxxxxxxxxxx>]

On 9/9/2022 6:48 PM, Kalle Valo wrote:
> Baochen Qiang <quic_bqiang@xxxxxxxxxxx> writes:
>
> > Currently this feature is enabled for QCA6390/WCN6855.
> >
> > Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1
> > Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-01720.1-QCAHSPSWPL_V1_V2_SILICONZ_LITE-1
> >
> > Signed-off-by: Baochen Qiang <quic_bqiang@xxxxxxxxxxx>
> I did quite a few changes to this patch in the pending branch, please
> check my changes:
>
> https://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git/commit/?h=pending&id=074477aacb419493da6fb4d96fa9d12390c3b40e
>
> I improved the commit log.
>
> > --- a/drivers/net/wireless/ath/ath11k/hw.h
> > +++ b/drivers/net/wireless/ath/ath11k/hw.h
> > @@ -126,6 +126,11 @@ struct ath11k_hw_hal_params {
> >   	enum hal_rx_buf_return_buf_manager rx_buf_rbm;
> >   };
> >   
> > +struct ath11k_hw_sram_dump {
> > +	u32 start;
> > +	u32 end;
> > +};
> > +
> >   struct ath11k_hw_params {
> >   	const char *name;
> >   	u16 hw_rev;
> > @@ -200,6 +205,7 @@ struct ath11k_hw_params {
> >   	bool hybrid_bus_type;
> >   	bool fixed_fw_mem;
> >   	bool support_off_channel_tx;
> > +	const struct ath11k_hw_sram_dump *sram_dump;
> >   };
> Instead of separate structures I used inline structures:
>
> 		.sram_dump = {
> 			.start = 0x01400000,
> 			.end = 0x0177ffff,
> 		},
>
> > --- a/drivers/net/wireless/ath/ath11k/pcic.c
> > +++ b/drivers/net/wireless/ath/ath11k/pcic.c
> > @@ -203,6 +203,35 @@ u32 ath11k_pcic_read32(struct ath11k_base *ab, u32 offset)
> >   }
> >   EXPORT_SYMBOL(ath11k_pcic_read32);
> >   
> > +int ath11k_pcic_dump_sram(struct ath11k_base *ab, u8 *buf,
> > +			  u32 start, u32 end)
> > +{
> > +	int ret = 0;
> > +	bool wakeup_required;
> > +	u32 *data = (u32 *)buf;
> I changed buf to a void pointer, then the cast is not needed.
>
> > +	u32 i;
> > +
> > +	/* for offset beyond BAR + 4K - 32, may
> > +	 * need to wakeup the device to access.
> > +	 */
> > +	wakeup_required = test_bit(ATH11K_FLAG_DEVICE_INIT_DONE, &ab->dev_flags) &&
> > +			  end >= ATH11K_PCI_ACCESS_ALWAYS_OFF;
> > +	if (wakeup_required && ab->pci.ops->wakeup) {
> > +		ret = ab->pci.ops->wakeup(ab);
> > +		if (ret)
> > +			ath11k_warn(ab, "%s: failed to do wakeup: %d\n", __func__, ret);
> > +	}
> I changed the error handling so that if wakeup() fails we do not
> continue and just return an error.

I prefer to keep the original design, because in that case we still have 
something to check after firmware crashes.

I admit that the dump content may be invalid if wakeup fails, but we can 
know that by checking kernel log, so we can avoid misleading.

> > +	for (i = start; i < end + 1; i += 4)
> > +		*data++ = ath11k_pcic_do_read32(ab, i);
> > +
> > +	if (wakeup_required && !ret && ab->pci.ops->release)
> > +		ab->pci.ops->release(ab);
> At the same time I removed the ret check here.
>
> > +
> > +	return 0;
> > +}
> > +EXPORT_SYMBOL(ath11k_pcic_dump_sram);
> I renamed this to ath11k_pcic_read() as I feel it's more descriptive
> what the function really does. It's not really care is this for sram
> dump or something else.
>
> I also renamed hif.h interface to ath11k_hif_read().