Date:   Fri, 03 Dec 2021 14:41:37 +0100
From:   Bjørn Mork <bjorn@...k.no>
To:     "Russell King (Oracle)" <linux@...linux.org.uk>
Cc:     Jakub Kicinski <kuba@...nel.org>, netdev@...r.kernel.org,
        照山周一郎 <teruyama@...ingboard-inc.jp>
Subject: Re: [PATCH net,stable] phy: sfp: fix high power modules without
 diag mode

"Russell King (Oracle)" <linux@...linux.org.uk> writes:

> On Thu, Dec 02, 2021 at 05:58:43PM -0800, Jakub Kicinski wrote:
>> On Tue, 30 Nov 2021 08:39:29 +0100 Bjørn Mork wrote:
>> > Commit 7cfa9c92d0a3 ("net: sfp: avoid power switch on address-change
>> > modules") changed semantics for high power modules without diag mode.
>> > We repeatedly try to read the current power status from the non-existing
>> > 0xa2 address, in the futile hope this failure is temporary:
>> > 
>> > [    8.856051] sfp sfp-eth3: module NTT              0000000000000000 rev 0000 sn 0000000000000000 dc 160408
>> > [    8.865843] mvpp2 f4000000.ethernet eth3: switched to inband/1000base-x link mode
>> > [    8.873469] sfp sfp-eth3: Failed to read EEPROM: -5
>> > [    8.983251] sfp sfp-eth3: Failed to read EEPROM: -5
>> > [    9.103250] sfp sfp-eth3: Failed to read EEPROM: -5
>> > 
>> > Eeprom dump:
>> > 
>> > 0x0000: 03 04 01 00 00 00 80 00 00 00 00 01 0d 00 0a 64
>> > 0x0010: 00 00 00 00 4e 54 54 20 20 20 20 20 20 20 20 20
>> > 0x0020: 20 20 20 20 00 00 00 00 30 30 30 30 30 30 30 30
>> > 0x0030: 30 30 30 30 30 30 30 30 30 30 30 30 05 1e 00 7d
>> > 0x0040: 02 00 00 00 30 30 30 30 30 30 30 30 30 30 30 30
>> > 0x0050: 30 30 30 30 31 36 30 34 30 38 20 20 00 00 00 75
>> > 0x0060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 0x0070: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 0x0080: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 0x0090: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 0x00a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 0x00b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 0x00c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 0x00d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 0x00e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 0x00f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>> > 
>> > Previously we assumed such modules were powered up in the correct
>> > mode, continuing without further configuration as long as the
>> > required power class was supported by the host.
>> > 
>> > Revert to that behaviour, refactoring to keep the improved
>> > diagnostic messages.
>> > 
>> > Fixes: 7cfa9c92d0a3 ("net: sfp: avoid power switch on address-change modules")
>> > Reported-and-tested-by: 照山周一郎 <teruyama@...ingboard-inc.jp>
>> > Cc: Russell King <rmk+kernel@...linux.org.uk>
>> > Signed-off-by: Bjørn Mork <bjorn@...k.no>
>> 
>> Russell, any comments?
>
> Sorry for the delay, I've been out over the last couple of days. I 
> hink it's fine, but the code here is not easy to understand, hence
> why this subtlety was missed. So, I'm not entirely happy about going
> back to the original code.
>
> Maybe instead doing a check in sfp_sm_mod_hpower() for this would
> be better? Possibly something like:
>
> static int sfp_module_parse_power(struct sfp *sfp)

You lost me now.  This is still changing sfp_module_parse_power() and
not sfp_sm_mod_hpower().  


> {
> 	u32 power_mW = 1000;
> +	bool supports_a2;
>
> 	if (sfp->id.ext.options & cpu_to_be16(SFP_OPTIONS_POWER_DECL))
> 		power_mW = 1500;
> 	if (sfp->id.ext.options & cpu_to_be16(SFP_OPTIONS_HIGH_POWER_LEVEL))
> 		power_mW = 2000;
>
> +	supports_a2 = sfp->id.ext.sff8472_compliance !=
> +				SFP_SFF8472_COMPLIANCE_NONE ||
> +			sfp->id.ext.diagmon & SFP_DIAGMON_DDM;
>
> 	if (power_mW > sfp->max_power_mW) {
> 		/* Module power specification exceeds the allowed maximum. */
> -		if (sfp->id.ext.sff8472_compliance ==
> -			SFP_SFF8472_COMPLIANCE_NONE &&
> -		    !(sfp->id.ext.diagmon & SFP_DIAGMON_DDM)) {
> +		if (!supports_a2) {
> ...
> 	}
> +
> +	if (!supports_a2 && power_mW > 1000) {
> +		/* The module power level is below the host maximum and the
> +		 * module appears not to implement bus address 0xa2, so assume
> +		 * that the module powers up in the indicated mode.
> +		 */
> +		return 0;
> +	}
>
> 	/* If the module requires a higher power mode, but also requires
> ...

Fine with me.  Looks cleaner, and should solve the reported problem
AFAICS.

> This way, if the module reports it doesn't support 0xa2, we don't get
> the "Address Change Sequence not supported" message - since if 0xa2 is
> not supported, then the address change sequence is irrelevant. However,
> modules shouldn't have that bit set... but "shouldn't" doesn't mean
> they do not.

Makes sense.  Although that's not a problem here, I guess we have to
expect just about any combination of random bits ;-)

> This also has the advantage of making the check explicit and obvious,
> and I much prefer the organisation of:
>
> 	if (module_exceeds_host_power) {
> 		handle this case
> 	} else {
> 		do other checks
> 	}


OK

> I think maybe dealing with power_mW <= 1000 early on may be a good idea,
> and eliminates the tests further down for power_mW > 1000.
>
> 	if (power_mW <= 1000) {
> 		sfp->module_power_mW = power_mW;
> 		return 0;
> 	}
>
> since those modules do not require any special handling.

Do you want this included now, or is that for a later cleanup?




Bjørn

Powered by blists - more mailing lists