Re: rtl8821ae keep alive not set, connection lost

[James Cameron <quozl@laptop.org>]

On Wed, Sep 20, 2017 at 04:48:23PM -0500, Larry Finger wrote:
> On 09/20/2017 04:36 AM, James Cameron wrote:
> >When the problem occurs, register 0x350 bit 25 is set, for which a
> >comment in _rtl8821ae_check_pcie_dma_hang says means there is an RX
> >hang.
> >
> >So perhaps driver should call _rtl8821ae_check_pcie_dma_hang
> >and _rtl8821ae_reset_pcie_interface_dma.
> >
> >Any ideas where to do this?
> 
> Thanks for the extended debugging.
> 
> I was able to repeat your findings. With the 8-bit read of
> REG_DBI_RDATA, I got poor connection stability. Reverting that part
> made it stable again. For that reason, I pushed the partial
> reversion of commit 40b368af4b75 ("rtlwifi: Fix alignment issues").

That's great you were able to reproduce, thanks!

> Where did you detect that bit 25 of register 0x350 was set?

In _rtl8821ae_check_pcie_dma_hang on link up.

REG_DBI_FLAG (0x350 bits 16-31) is observed as;

- 0x0000 on entry to function after warm boot,

- 0x0400 on exit from function; debug bit 23 is set by the function,

- 0x0400 on entry to function on link up when the problem has not
  happened,

- 0x0600 on entry to function on link up when the problem has
  happened.

But I don't know if 0x0600 is useful to detect earlier, or if it is
only a symptom of link down while device active.  Either way, if it
truly does signal an RX hang or firmware RX queue full, it's useful.

My "-q9" and "-qa" test kernels dump REG_DBI_CTRL and REG_DBI_FLAG.

"-q9" is with 8-bit read of REG_DBI_RDATA.

"-qa" is with 16-bit read of REG_DBI_DATA.

My "-qa" test kernel;
http://dev.laptop.org/~quozl/y/1dunwN.txt (git diff v4.13)
http://dev.laptop.org/~quozl/z/1dubX7.txt (dmesg)

REG_DBI_CTRL+3 used by _rtl8821ae_check_pcie_dma_hang is effectively
REG_DBI_FLAG+1 (0x353).

REG_DBI_CTRL is REG_DBI_ADDR; a duplicate register definition.

I'm still pondering a few more theories;

- change write_readback, it is true now, and the while()/udelay in
  _rtl8821ae_dbi_read seems a waste, it never executes,

- clearing REG_DBI_CTRL write enable bits at the end of
  _rtl8821ae_dbi_write,

- switching to 32-bit access as used by rtl8192de.

And a giggle from reviewing the code,
_rtl8821ae_wowlan_initialize_adapter says "Patch Pcie Rx DMA hang
after S3/S4 several times.  The root cause has not be found."
... I've learned that root causes that aren't found tend to cause
further problems later.  ;-)

Given this, my gut feel is firmware or silicon problem; RX DMA ceases,
the driver does not detect it, the connection is lost.

-- 
James Cameron
http://quozl.netrek.org/