Re: [PATCH v6 03/24] rtw89: add core and trx files

[Kalle Valo <kvalo@codeaurora.org>]

Pkshih <pkshih@realtek.com> writes:

>> -----Original Message-----
>> From: kvalo=codeaurora.org@mg.codeaurora.org
>> <kvalo=codeaurora.org@mg.codeaurora.org> On
>> Behalf Of Kalle Valo
>> Sent: Saturday, October 2, 2021 12:26 AM
>> To: Pkshih <pkshih@realtek.com>
>> Cc: linux-wireless@vger.kernel.org
>> Subject: Re: [PATCH v6 03/24] rtw89: add core and trx files
>> 
>> Ping-Ke Shih <pkshih@realtek.com> writes:
>> 
>> > Implement main flows that contains register/unregister mac80211 hw with
>> > hardware capability, power on/off sequence, STA state actions, and
>> > TX/RX path.
>> >
>> > The chip info is read from efuse while probing PCI, and then it can be
>> > used to induce supported channel, band, bitrate, ht/vht/he capability,
>> > and etc. Then, we register hardware with these capabilities.
>> >
>> > When network interface is up, driver does power-on sequence to enable MAC,
>> > BB and RF function blocks. Oppositely, do power-off sequence when
>> > interface is going to down.
>> >
>> > To maintain STA state, five callbacks are implemented -- add, assoc,
>> > disassoc, disconnect and remove. In which state, driver tells firmware STA
>> > info via H2C.
>> >
>> > TX flow:
>> > When a SKB is going to be transmitted, we must know its type first. If
>> > the type is mgmt or fwcmd made by driver, SKB is queued into corresponding
>> > DMA channel and PCI ring. The other type is data frame that is more
>> > complex, because it needs to establish BA session to have better throughput
>> > with AMPDU and AMSDU.
>> > In order to have better PCI DMA efficiency, we don't kick off DMA every
>> > SKB. With wake TX queue, kick off DMA after a bunch of SKBs are written.
>> > To achieve this, we have two HCI ops -- tx_write and tx_kick_off.
>> >
>> > BA establishment work:
>> > For data frames, we start to establish BA session if the STA is associated
>> > with APMDU capability and the TID session isn't established, and then the
>> > BA work is used to ask mac80211 to start AMPDU actions. Driver implements
>> > AMPDU action callbacks to know the session is established, so that we can
>> > set AGG_EN bit in TX descriptor to enable AMPDU.
>> >
>> > RX flow:
>> > When a RX SKB is delivered from PCI, rtw89_core_rx() process it depneds on
>> > its type -- WIFI, C2H or PPDU. If type is C2H, it's queued into a C2H
>> > queue, and wake a work to handle the C2H packet. If type is WIFI, it's a
>> > normal RX packet. When mgmt or data frame is received, it is queued
>> > into pending RX SKB queue to wait for corresponding PPDU packet (another
>> > RX packet with PPDU type) to fill its rx_status, like RSSI. And, then
>> > indicate this packet to mac80211. When control frame is received, indicate
>> > it to mac80211 immediately.
>> >
>> > Track work:
>> > Use track work to monitor PHY status to know the changes of environment,
>> > and then update RA status or do RFK accordingly.
>> >
>> > Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
>> 
>> [...]
>> 
>> > +static __always_inline void RTW89_SET_TXWD(u8 *txdesc, u32 val,
>> > u8 offset, u32 mask)
>> > +{
>> > +	u32 *txd32 = (u32 *)txdesc;
>> > +
>> > +	le32p_replace_bits((__le32 *)(txd32 + offset), val, mask);
>> > +}
>> 
>> I'm not convinced about this either, please just use inline.
>
> This is because 'mask' argument of le32p_replace_bits() must be constant
> only. If I use inline and build this driver with ccflags-y += -Os,
> compiler reports errors:
>
> In function 'field_multiplier',
>     inlined from 'le32_encode_bits' at ./include/linux/bitfield.h:154:1,
>     inlined from 'le32p_replace_bits' at ./include/linux/bitfield.h:154:1,
>     inlined from 'RTW89_SET_FWCMD_UA32.constprop' at /work/git-root/rtwlan/rtw89/fw.h:1397:2:
> ./include/linux/bitfield.h:119:3: error: call to '__bad_mask' declared with attribute error: bad bitfield mask
>   119 |   __bad_mask();
>       |   ^~~~~~~~~~~~
>
> I check the implement of le32p_replace_bits(), it looks like
>
> static __always_inline void type##p_replace_bits(__##type *p,           \
>                                         base val, base field)           \
> {                                                                       \
>         *p = (*p & ~to(field)) | type##_encode_bits(val, field);        \
> }
>
> So, I imitate the function to use __always_inline, and then it works.
>
> Do you think I don't need to consider the case of Os?
> But, -Os seems a standard option of Linux kernel.
>
> ifdef CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE
> KBUILD_CFLAGS += -O2
> else ifdef CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE_O3
> KBUILD_CFLAGS += -O3
> else ifdef CONFIG_CC_OPTIMIZE_FOR_SIZE
> KBUILD_CFLAGS += -Os
> endif

Yeah, we need to support -Os.

Arnd, what do you suggest? Is __always_inline good solution for this? I
think we should at least add a comment explaining why it's needed.

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches