linux-kernel - Re: [PATCH] net: phy: broadcom: add support for BCM5221 phy

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <4a67a6ae-f697-1353-1b0f-55b42e5dbdfa@benettiengineering.com>
Date:   Sat, 12 Aug 2023 19:27:56 +0200
From:   Giulio Benetti <giulio.benetti@...ettiengineering.com>
To:     Florian Fainelli <florian.fainelli@...adcom.com>,
        Broadcom internal kernel review list 
        <bcm-kernel-feedback-list@...adcom.com>,
        Andrew Lunn <andrew@...n.ch>,
        Heiner Kallweit <hkallweit1@...il.com>,
        "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>
Cc:     Russell King <linux@...linux.org.uk>, netdev@...r.kernel.org,
        linux-kernel@...r.kernel.org, Jim Reinhart <jimr@...vox.com>,
        James Autry <jautry@...vox.com>,
        Matthew Maron <matthewm@...vox.com>
Subject: Re: [PATCH] net: phy: broadcom: add support for BCM5221 phy

Hello Florian,

thanks for reviewing,

On 12/08/23 01:52, Florian Fainelli wrote:
> 
> 
> On 8/11/2023 2:53 PM, Giulio Benetti wrote:
>> This patch adds the BCM5221 PHY support by reusing
>> brcm_fet_config_intr() and brcm_fet_handle_interrupt() and
>> implementing config_init()/suspend()/resume().
>>
>> Sponsored by: Tekvox Inc.
>> Cc: Jim Reinhart <jimr@...vox.com>
>> Cc: James Autry <jautry@...vox.com>
>> Cc: Matthew Maron <matthewm@...vox.com>
>> Signed-off-by: Giulio Benetti <giulio.benetti@...ettiengineering.com>
> 
> Looks good, few comments below.
> 
>> ---
>>   drivers/net/phy/broadcom.c | 144 +++++++++++++++++++++++++++++++++++++
>>   include/linux/brcmphy.h    |   8 +++
>>   2 files changed, 152 insertions(+)
>>
>> diff --git a/drivers/net/phy/broadcom.c b/drivers/net/phy/broadcom.c
>> index 59cae0d808aa..99f6c0485f01 100644
>> --- a/drivers/net/phy/broadcom.c
>> +++ b/drivers/net/phy/broadcom.c
>> @@ -754,6 +754,84 @@ static int brcm_fet_config_init(struct phy_device 
>> *phydev)
>>       return err;
>>   }
>> +static int bcm5221_config_init(struct phy_device *phydev)
>> +{
> 
> Very similar to brcm_fet_config_init() except that you configure fewer 
> interrupt sources, do not have the LED mode programming and your MDI-X 
> programming is done via a 10BaseT specific register rather than the shadow.
> 
> Can you consider adding parameters to brcm_fet_config_init() such that 
> you can specify the 5221 specific settings such as the interrupt mask 
> for instance?

So if I've understood correctly I should drop bcm5221_config_init() and
use brcm_fet_config_init() by checking the struct phy_driver .phy_id to
be PHY_ID_BCM5221.

> This should help address the locking that Russell suggested.

But I don't understand how this can help to prevent the locking Russell
suggested, can you please elaborate more or point me and example?
As far as I've understood I should take care to lock sections to prevent
from interrupts to occur during shadow mode switching, is it correct?
So maybe you mean that if I do this in brcm_fet_config_init() it will be
fixed for the other phys using brcm_fet_config_init(), correct?

> [snip]
>> +static int bcm5221_suspend(struct phy_device *phydev)
>> +{
>> +    int reg, err, err2, brcmtest;
>> +
>> +    /* Enable shadow register access */
>> +    brcmtest = phy_read(phydev, MII_BRCM_FET_BRCMTEST);
>> +    if (brcmtest < 0)
>> +        return brcmtest;
>> +
>> +    reg = brcmtest | MII_BRCM_FET_BT_SRE;
>> +
>> +    err = phy_write(phydev, MII_BRCM_FET_BRCMTEST, reg);
>> +    if (err < 0)
>> +        return err;
>> +
>> +    /* Force Low Power Mode with clock enabled */
>> +    err = phy_set_bits(phydev, MII_BRCM_FET_SHDW_AUXMODE4,
>> +               BCM5221_SHDW_AM4_EN_CLK_LPM |
>> +               BCM5221_SHDW_AM4_FORCE_LPM);
>> +
>> +    /* Disable shadow register access */
>> +    err2 = phy_write(phydev, MII_BRCM_FET_BRCMTEST, brcmtest);
>> +    if (!err)
>> +        err = err2;
>> +
>> +    return err;
>> +}
> 
> bcm5221_suspend() is essentially brcm_fet_suspend() minus the setting of 
> the BMCR.PDOWN bit, can you consider factoring brcm_fet_suspend() into a 
> helper that can decide whether to set the power down bit or not? Does 
> not brcm_fet_suspend() work as-is?

Sure, I will,

>> +
>> +static int bcm5221_resume(struct phy_device *phydev)
>> +{
> 
> This should really be calling bcm5221_config_init() here, if the PHY was 
> on a power island that is powered off during system suspend, 
> bcm5221_resume() would not be restoring the auto-power down that is set 
> during config_init(), probably not desired because that means you will 
> burn power when the cable is disconnected...

oh yes! You're right, I will then drop bcm5221_resume() in favor of 
bcm5221_config_init(), so brcm_fet_config_init() with phy_id == 
PHY_ID_BCM5221 checks.

Thanks again
Best regards
-- 
Giulio Benetti
CEO&CTO@...etti Engineering sas