Date: Wed, 13 Sep 2023 15:26:25 +0200
From: Lukasz Majewski <lukma@...x.de>
To: Andrew Lunn <andrew@...n.ch>
Cc: Parthiban Veerasooran <Parthiban.Veerasooran@...rochip.com>,
 davem@...emloft.net, edumazet@...gle.com, kuba@...nel.org,
 pabeni@...hat.com, robh+dt@...nel.org, krzysztof.kozlowski+dt@...aro.org,
 conor+dt@...nel.org, corbet@....net, steen.hegelund@...rochip.com,
 rdunlap@...radead.org, horms@...nel.org, casper.casan@...il.com,
 netdev@...r.kernel.org, devicetree@...r.kernel.org,
 linux-kernel@...r.kernel.org, linux-doc@...r.kernel.org,
 horatiu.vultur@...rochip.com, Woojung.Huh@...rochip.com,
 Nicolas.Ferre@...rochip.com, UNGLinuxDriver@...rochip.com,
 Thorsten.Kummermehr@...rochip.com
Subject: Re: [RFC PATCH net-next 2/6] net: ethernet: add mac-phy interrupt
 support with reset complete handling

Hi Andrew,

> > Just maybe mine small remark. IMHO the reset shall not pollute the
> > IRQ hander. The RESETC is just set on the initialization phase and
> > only then shall be served. Please correct me if I'm wrong, but it
> > will not be handled during "normal" operation.  
> 
> This is something i also wondered. Maybe if the firmware in the
> MAC-PHY crashes, burns, and a watchdog reset it, could it assert
> RESETC? I think maybe a WARN_ON_ONCE() for RESETC in the interrupt
> handler would be useful, but otherwise ignore it. Probe can then poll
> during its reset.
> 
> > > +				regval = RESETC;
> > > +				/* SPI host should write RESETC
> > > bit with one to
> > > +				 * clear the reset interrupt
> > > status.
> > > +				 */
> > > +				ret = oa_tc6_perform_ctrl(tc6,
> > > OA_TC6_STS0,
> > > +
> > > &regval, 1, true,
> > > +
> > > false);  
> > 
> > Is this enough to have the IRQ_N deasserted (i.e. pulled HIGH)?
> > 
> > The documentation states it clearly that one also needs to set SYNC
> > bit (BIT(15)) in the OA_CONFIG0 register (which would have the
> > 0x8006 value).
> > 
> > Mine problem is that even after writing 0x40 to OA_STATUS0 and
> > 0x8006 to OA_CONFIG0 the IRQ_N is still LOW (it is pulled up via
> > 10K resistor).
> > 
> > (I'm able to read those registers and those show expected values)  
> 
> What does STATUS0 and STATUS1 contain?

STATUS0 => 0x40, which is expected.

Then I do write 0x40 to STATUS0 -> bit6 (RESETC) is R/W1C

After reading the same register - I do receive 0x00 (it has been
cleared).

Then I write 0x8006 to OA_CONFIG0.

(Those two steps are regarded as "configuration" of LAN865x device in
the documentation)

In this patch set -> the OA_COFIG0 also has the 0x6 added to indicate
the SPI transfer chunk.

Dump of OA registers:
{0x11, 0x7c1b3, 0x5e5, 0x0, 0x8006, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
0x3000, 0x1fbf, 0x3ffe0003, 0x0, 0x0}

Status 0 (0x8) -> 0x0
Status 1 (0x9) -> 0x0

> That might be a dumb question,
> i've not read the details for interrupt handling yet, but maybe there
> is another interrupt pending? Or the interrupt mask needs writing?

All the interrupts on MASK{01} are masked. 

Changing it to:
sts &= ~(OA_IMASK0_TXPEM | OA_IMASK0_TXBOEM | OA_IMASK0_TXBUEM |
OA_IMASK0_RXBOEM | OA_IMASK0_LOFEM | OA_IMASK0_HDREM

doesn't fix this problem.

> 
> > Was it on purpose to not use the RST_N pin to perform GPIO based
> > reset?
> > 
> > When I generate reset pulse (and keep it for low for > 5us) the
> > IRQ_N gets high. After some time it gets low (as expected). But
> > then it doesn't get high any more.  
> 
> Does the standard say RST_N is mandatory to be controlled by software?
> I could imagine RST_N is tied to the board global reset when the power
> supply is stable.

It can be GPIO controlled. However, it is not required. I've tied it to
3V3 and also left NC, but no change.

> Software reset is then used at probe time.

I've reconfigured the board to use only SW based reset (i.e. set bit0
in OA_RESET - 0x3).

> 
> So this could be a board design decision. I can see this code getting
> extended in the future, an optional gpiod passed to the core for it to
> use.

I can omit the RST_N control. I'd just expect the IRQ_N to be high
after reset.

> 
> > > msecs_to_jiffies(1));  
> > 
> > Please also clarify - does the LAN8651 require up to 1ms "settle
> > down" (after reset) time before it gets operational again?  
> 
> If this is not part of the standard, it really should be in the MAC
> driver, or configurable, since different devices might need different
> delays. But ideally, if the status bit says it is good to go, i would
> really expect it to be good to go. So this probably should be a
> LAN8651 quirk.

The documentation is silent about the "settle down time". The only
requirements is for RST_N assertion > 5us. However, when the IRQ_N goes
low, and the interrupt is served - it happens that I cannot read ID
from the chip via SPI.

> 
> 	Andrew

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH,      Managing Director: Erika Unter
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma@...x.de

Content of type "application/pgp-signature" skipped

Powered by blists - more mailing lists