Thread (1 message) 1 message, 1 author, 2022-02-22

Re: [PATCH v2 net-next 2/2] net: dsa: realtek: rtl8365mb: serialize indirect PHY register access

From: Alvin Šipraga <hidden>
Date: 2022-02-22 16:06:19
Also in: lkml

Linus Walleij [off-list ref] writes:
On Tue, Feb 22, 2022 at 1:19 AM Alvin Šipraga [off-list ref] wrote:
quoted
quoted
On Mon, Feb 21, 2022 at 7:46 PM Alvin Šipraga [off-list ref] wrote:
quoted
quoted
What I worry about is whether we need to do a similar patch to
rtl8366rb_phy_read() and rtl8366rb_phy_write() in
rtl8366rb.c?
Unfortunately I do not have the hardware to test rtl8366rb.c, so I can
only speculate. But I gave some hints in the commit message which might
help in checking whether or not it is an issue on that hardware as
well. The code for rtl8366rb_phy_read() looks similar, but since this is
a quirk of the hardware design, it could be that it is not
necessary. The only way is to test.
I can test it.
Great! I have attached a test patch to insert these stray register
accesses inside the PHY read/write functions:

0001-TEST-net-dsa-realtek-rtl8366rb-provoke-stray-reg-acc.patch

Set the module parameter stray_access to 1 in order to enable stray
register access. I did this in case you don't build it as a module but
still want to do A/B testing - just remove it if I goofed.
quoted
If you or somebody else can confirm that it is an issue for RTL8366RB as
well, I will happily send a patch to the list for testing. You can for
example try spamming PHY register reads with phytool while also reading
off switch registers via regmap debugfs. See also the discussion in [1].

[1] https://lore.kernel.org/netdev/878rukib4f.fsf@bang-olufsen.dk/ (local)
If you have time to write a patch I'd love testing it!
Attached also a patch with a proposed fix - if indeed it is an issue -
along the same lines as the second patch in this series. You'll also
need the first patch from this series (attached for convenience):

0001-net-dsa-realtek-allow-subdrivers-to-externally-lock-.patch
0002-TEST-net-dsa-realtek-rtl8366rb-serialize-indirect-PH.patch

All patches are against net-next and build-tested only, so you might
want to double check if the results are unexpected.

Kind regards,
Alvin
Yours,
Linus Walleij
  

Attachments

Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help