linux-kernel - Re: [PATCH v1] ALSA: hda/tas2781: Prevent throwing away custom calibration addresses

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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <a74b90ee-5885-4acc-88f0-60642e0284e5@linux.dev>
Date: Wed, 7 Jan 2026 19:29:11 -0800
From: Matthew Schwartz <matthew.schwartz@...ux.dev>
To: Antheas Kapenekakis <lkml@...heas.dev>
Cc: Shenghao Ding <shenghao-ding@...com>, Baojun Xu <baojun.xu@...com>,
 tiwai@...e.de, linux-kernel@...r.kernel.org, linux-sound@...r.kernel.org
Subject: Re: [PATCH v1] ALSA: hda/tas2781: Prevent throwing away custom
 calibration addresses

On 1/7/26 2:21 AM, Antheas Kapenekakis wrote:
> On Wed, 7 Jan 2026 at 12:17, Matthew Schwartz
> <matthew.schwartz@...ux.dev> wrote:
>>
>>
>>
>>> On Jan 7, 2026, at 1:32 AM, Antheas Kapenekakis <lkml@...heas.dev> wrote:
>>>
>>> The V2 UEFI spec for per-device calibration data in TAS allows for
>>> manufacturers to specify up to five custom calibration register
>>> addresses and values to optimize per-device in factory.
>>>
>>> A basic description of the spec is found in a comment in function
>>> tas2781_apply_calib(). However, currently, for the condition of
>>> p->dspbin_typ != TASDEV_BASIC, if custom calibration addresses have been
>>> provided by the manufacturer, they are ignored and the ones listed in
>>> firmware are used.
>>>
>>> In case the manufacturer chose to change different registers other than
>>> the default ones (r0, invr0, r0_low, pow, tlimit), this causes the
>>> UEFI calibration data to scratch different registers and cause the
>>> firmware to misbehave. This is true in the Xbox ROG Ally X, for one of
>>> the firmwares, where it pops and has audio dropouts after the
>>> calibration data is applied.
>>>
>>> Therefore, add a new bool to indicate that UEFI supplied custom
>>> calibration addresses, and if so, use them regardless of dspbin_typ.
>>>
>>> Link: https://lore.kernel.org/all/CAGwozwFQKoQgo_Q=qdr-FTD+uoVWA8AtHyDmKwTOV4ZU6+F3SQ@mail.gmail.com/
>>> Reported-by: Matthew Schwartz <matthew.schwartz@...ux.dev>
>>> Closes: https://lore.kernel.org/all/0ba100d0-9b6f-4a3b-bffa-61abe1b46cd5@linux.dev/
>>> Co-developed-by: Matthew Schwartz <matthew.schwartz@...ux.dev>
>>> Signed-off-by: Matthew Schwartz <matthew.schwartz@...ux.dev>
>>> Signed-off-by: Antheas Kapenekakis <lkml@...heas.dev>
>>>
>>> ---
>>> @Matt: can you kindly verify this patch works for you?
>>
>> I’ll have time to debug it more in the morning but the patch does not work, it’s still using the same calibrated data as before with dropouts.
> 
> If you can check that the register addresses changed that would be
> great. Especially that we enter and exit the calibration block.
> Perhaps that was happening before as well and the addresses were
> overwritten by firmware later

[    2.415699] snd_hda_codec_alc269 hdaudioC1D0: bound i2c-TXNW2781:00-tas2781-hda.0 (ops tas2781_hda_comp_ops [snd_hda_scodec_tas2781_i2c])
[    2.417548] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: dspbin_type_check: ppcver=0x19500 -> dspbin_typ=2 (0=none,1=BASIC,2=ALPHA,3=BETA)
[    3.809849] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tas2781_apply_calib: enter, dspbin_typ=2, chip_id=0xadd
[    3.809855] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tas2781_apply_calib: default cali_reg: r0=0xbf4 invr0=0xc0c r0_low=0xc14 pow=0x9f0 tlimit=0xc7c
[    3.809858] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tas2781_apply_calib: UEFI custom cali_reg found: r0=0xce4 invr0=0xcf4 r0_low=0xcfc pow=0xae0 tlimit=0xd70
[    3.809860] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tas2781_apply_calib: using cali_reg for r->regs (custom_addr=1, dspbin_typ=2)
[    3.809861] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tas2781_apply_calib: exit, total_sz=42, is_user_space_calidata=1
[    6.180351] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tas2781_hda_playback_hook: action = 0
[    6.330017] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tasdev_load_calibrated_data: enter dev 0, is_user_space_calidata=1
[    6.330021] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tasdev_load_calibrated_data: writing UEFI cal data to regs: r0=0xce4 invr0=0xcf4 r0_low=0xcfc pow=0xae0 tlimit=0xd70
[    6.330022] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tasdev_load_calibrated_data: cal values: r0=3c5f7222 invr0=10f615d5
[    6.332137] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tasdev_load_calibrated_data: enter dev 1, is_user_space_calidata=1
[    6.332138] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tasdev_load_calibrated_data: writing UEFI cal data to regs: r0=0xce4 invr0=0xcf4 r0_low=0xcfc pow=0xae0 tlimit=0xd70
[    6.332138] tas2781-hda i2c-TXNW2781:00-tas2781-hda.0: tasdev_load_calibrated_data: cal values: r0=3c90e72d invr0=10e83c28

Looks like the addresses changed. I'm just going to send out a patch to skip calibration for the device as Jim suggested, as that is confirmed to work.

> 
> Antheas
> 
>>>
>>> Then verify your sign-off for the coby tag. If you would rather a
>>> different acknowledgment tag (e.g., tested-by or suggested-by), lmk
>>>
>>> Thanks.
>>> Antheas
>>> ---
>>> sound/hda/codecs/side-codecs/tas2781_hda.c | 4 +++-
>>> 1 file changed, 3 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/sound/hda/codecs/side-codecs/tas2781_hda.c b/sound/hda/codecs/side-codecs/tas2781_hda.c
>>> index 96e6d82dc69e..6efdf0a3442b 100644
>>> --- a/sound/hda/codecs/side-codecs/tas2781_hda.c
>>> +++ b/sound/hda/codecs/side-codecs/tas2781_hda.c
>>> @@ -64,6 +64,7 @@ static void tas2781_apply_calib(struct tasdevice_priv *p)
>>>       TASDEVICE_REG(0, 0x18, 0x7c),
>>>   };
>>>   unsigned int crc, oft, node_num;
>>> +    bool custom_addr = false;
>>>   unsigned char *buf;
>>>   int i, j, k, l;
>>>
>>> @@ -104,6 +105,7 @@ static void tas2781_apply_calib(struct tasdevice_priv *p)
>>>       for (j = 0, k = 0; j < node_num; j++) {
>>>           oft = j * 6 + 3;
>>>           if (tmp_val[oft] == TASDEV_UEFI_CALI_REG_ADDR_FLG) {
>>> +                custom_addr = true;
>>>               for (i = 0; i < TASDEV_CALIB_N; i++) {
>>>                   buf = &data[(oft + i + 1) * 4];
>>>                   cali_reg[i] = TASDEVICE_REG(buf[1],
>>> @@ -151,7 +153,7 @@ static void tas2781_apply_calib(struct tasdevice_priv *p)
>>>       }
>>>   }
>>>
>>> -    if (p->dspbin_typ == TASDEV_BASIC) {
>>> +    if (custom_addr || p->dspbin_typ == TASDEV_BASIC) {
>>>       r->r0_reg = cali_reg[0];
>>>       r->invr0_reg = cali_reg[1];
>>>       r->r0_low_reg = cali_reg[2];
>>>
>>> base-commit: 9ace4753a5202b02191d54e9fdf7f9e3d02b85eb
>>> --
>>> 2.52.0
>>
>