On 11/08/2017 03:15 AM, Russell King - ARM Linux wrote:
> On Tue, Nov 07, 2017 at 07:49:08PM -0800, Florian Fainelli wrote:
>> The SFP module identification code in sfp_sm_mod_probe() will reject SFF
>> modules soldered down because they have an identified of 0x2, while the code
>> currently checks for 0x3 only (SFP_PHYS_ID_SFP), update that.
>>
>> Signed-off-by: Florian Fainelli <f.faine...@gmail.com>
>> ---
>> drivers/net/phy/sfp.c | 5 +++--
>> include/linux/sfp.h | 1 +
>> 2 files changed, 4 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/net/phy/sfp.c b/drivers/net/phy/sfp.c
>> index e381811e5f11..942288aa9cdb 100644
>> --- a/drivers/net/phy/sfp.c
>> +++ b/drivers/net/phy/sfp.c
>> @@ -463,8 +463,9 @@ static int sfp_sm_mod_probe(struct sfp *sfp)
>> vendor, part, rev, sn, date);
>>
>> /* We only support SFP modules, not the legacy GBIC modules. */
>> - if (sfp->id.base.phys_id != SFP_PHYS_ID_SFP ||
>> - sfp->id.base.phys_ext_id != SFP_PHYS_EXT_ID_SFP) {
>> + if ((sfp->id.base.phys_id != SFP_PHYS_ID_SFP &&
>> + sfp->id.base.phys_id != SFP_PHYS_ID_SFF) ||
>> + sfp->id.base.phys_ext_id != SFP_PHYS_EXT_ID_SFP) {
>
> I'd prefer that we do something like the patch I sent a couple of nights
> ago, having a separate compatible for the SFF modules (since they have
> no insert signal as SFF is soldered in place) and use that to decide
> which phys_id we accept here.
Fair enough.
--
Florian
