Re: [BUG] wifi: rtw88: Hard system freeze on RTL8821CE when power_save is enabled (LPS/ASPM conflict)

[LB F]

Hi Ping-Ke,

I apologize for the rapid follow-up and for being perhaps a bit over-assertive
in my previous email. As I continued to dig into the code, I realized that
some of my interpretations of hardware registers (like REG_RX_PKT_LIMIT)
and kernel serialization might be simplified compared to the real-world
complexities you deal with.

I'd like to reframe my previous notes as "curious observations" that I
stumbled upon while testing, and I'd value your professional take on whether
they are relevant:

1. RX Host-Side Validation:
While searching for the 12KB limit I mentioned, I noticed that in
rtw_pci_rx_napi(), the driver uses the pkt_len field from the descriptor
directly for skb_put_data() without checking it against the host buffer
size (RTK_PCI_RX_BUF_SIZE). Even if the hardware normally clips DMA,
would it be worth adding a host-side guard there as a "hardening" measure
against potentially malformed hardware reports?

2. TX Write Pointer (wp) Fetch:
I noticed that in rtw_pci_tx_write_data(), get_tx_buffer_desc() fetches
the wp outside the irq_lock. I wasn't sure if mac80211 guarantees that
the direct TX path and the background worker threads can never collide on
the same queue, but I thought it was worth mentioning just in case.

3. Memory Barriers:
The wmb() point was more of an architectural observation regarding
PCI best practices for non-x86 platforms. I understand x86 is quite
forgiving here, but I noticed it was a pattern that stood out.

Please treat these as humble suggestions from someone trying to learn
the driver's internals. I didn't mean to imply these were "critical bugs"
without your expert verification.

Thank you for your patience with my technical excitement!

Best regards,
Oleksandr