Re: WARNING at drivers/acpi/platform_profile.c:74 in platform_profile_show()

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 25. 09. 23, 7:42, Jiri Slaby wrote:
On 25. 09. 23, 7:35, Jiri Slaby wrote:
Hi,

according to logs, since 6.3 (up to 6.5.4 now), I repeatedly see:
 > WARNING: CPU: 14 PID: 962 at drivers/acpi/platform_profile.c:74 platform_profile_show+0xb1/0x100 [platform_profile]
 > Modules linked in: ccm michael_mic ...
 > CPU: 14 PID: 962 Comm: power-profiles- Kdump: loaded Not tainted 6.5.4-6-default #1 openSUSE Tumbleweed (unreleased) dd37106c593be78644bb80e3c1534d801bf4cb36  > Hardware name: LENOVO 21CRS0K83K/21CRS0K83K, BIOS R22ET60W (1.30 ) 02/09/2023
 > RIP: 0010:platform_profile_show+0xb1/0x100 [platform_profile]
 > Code: d0 a8 ...
 > RSP: 0018:ffff9c1ac0b97db0 EFLAGS: 00010296
 > RAX: 0000000000000000 RBX: 0000000000000000 RCX: 000000008fc35be0
 > RDX: 0000000000000000 RSI: ffff9c1ac0b97db4 RDI: ffffffffc0a8b0a0
 > RBP: ffff8955ca540000 R08: ffff895b9f1ed180 R09: ffff895559ea1bc0
 > R10: 00000000031a400e R11: 000000000003f680 R12: ffff895b9f1ed180
 > R13: ffff9c1ac0b97e50 R14: 0000000000000001 R15: ffff9c1ac0b97ee8
 > FS:  00007f71b0e71900(0000) GS:ffff895b9f100000(0000) knlGS:0000000000000000
 > CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 > CR2: 00007fe402ea3400 CR3: 000000012004c000 CR4: 0000000000750ee0
 > PKRU: 55555554
 > Call Trace:
 >  <TASK>
 >  sysfs_kf_seq_show+0xab/0x100
 >  seq_read_iter+0x123/0x480
 >  vfs_read+0x1b8/0x300

It's:
WARN_ON((profile < 0) || (profile >= ARRAY_SIZE(profile_names))))

So I put there one more print:
dev_warn(dev, "profile=%d profile_get=%ps\n",
          profile, cur_profile->profile_get);

and I see:
: profile=-1883022368 profile_get=dytc_profile_get [thinkpad_acpi]
: profile=-1510173440 profile_get=dytc_profile_get [thinkpad_acpi]
: profile=-1510173440 profile_get=dytc_profile_get [thinkpad_acpi]
: profile=-966231712 profile_get=dytc_profile_get [thinkpad_acpi]
: profile=-1578420592 profile_get=dytc_profile_get [thinkpad_acpi]
: profile=-1578420592 profile_get=dytc_profile_get [thinkpad_acpi]
: profile=-1578420592 profile_get=dytc_profile_get [thinkpad_acpi]

where the profile values in hex are like:
ffffffff8fc35be0
ffffffffa5fc9500
ffffffffc6687960

Looking at simplicity of dytc_profile_get(), I wonder how that can happen.

I also wonder about dev passed to dytc_profile_get() having empty name (nothing before colon above)? Is that expected?

I forgot to add:
   cat /sys/firmware/acpi/platform_profile
correctly returns:
   low-power

without any WARNING. It looks like the warn often (but not exclusively) happens around:
PM: suspend exit

Any ideas?

Ah, convert_dytc_to_profile()'s retval is not checked in dytc_profile_refresh(). That's likely it, let me add the check and retry.

OK,
so adding:
--- a/drivers/platform/x86/thinkpad_acpi.c
+++ b/drivers/platform/x86/thinkpad_acpi.c
@@ -10418,7 +10418,14 @@ static void dytc_profile_refresh(void)
                return;

        perfmode = (output >> DYTC_GET_MODE_BIT) & 0xF;
-       convert_dytc_to_profile(funcmode, perfmode, &profile);
+       err = convert_dytc_to_profile(funcmode, perfmode, &profile);
+       if (err) {
+ pr_warn("%s: mmc=%u psc=%u mmc_get=%u funcmode=%d output=0x%x perfmode=%d\n", + __func__, !!(dytc_capabilities & BIT(DYTC_FC_MMC)),
+                          !!(dytc_capabilities & BIT(DYTC_FC_PSC)),
+ dytc_mmc_get_available, funcmode, output, perfmode);
+               return;
+       }
        if (profile != dytc_current_profile) {
                dytc_current_profile = profile;
                platform_profile_notify();

fixes the warning, of course.

The output is:
dytc_profile_refresh: mmc=0 psc=1 mmc_get=0 funcmode=0 output=0x1f001 perfmode=15

So using psc mode, it retrieves DYTC (0x1f001) as follows:
0x1.... -> VSTD=1 -> STD mode
0x.f... -> CICM=0xf, DYTC_GET_MODE_BIT(12) -> dunno what it is in STD
0x..0.. -> CICF=0x0, DYTC_GET_FUNCTION_BIT(8) -> DYTC_FUNCTION_STD
0x....1 -> GOOD

But convert_dytc_to_profile() doesn't handle this at all. Do I have a newer DYTC interface? Or a broken one?

My DYTC for CMD_GET looks like:
   Case (0x02)
   {
       Local5 = VSTD /* \VSTD */
       Local5 |= (VCQL << 0x01)
       Local5 |= (VSTP << 0x04)
       Local5 |= (VADM << 0x07)
       Local5 |= (VTMS << 0x09)
       Local5 |= (VDLS << 0x0A)
       Local5 |= (VMSC << 0x0C)
       Local5 |= (VPSC << 0x0D)
       Local1 = (CICF << 0x08)
       If ((CICF == 0x03))
       {
           CICM = SMYH /* \SMYH */
       }
       ElseIf ((CICF == 0x0B))
       {
           CICM = SMMC /* \SMMC */
       }
       ElseIf ((CICF == 0x0D))
       {
           CICM = SPSC /* \SPSC */
       }
       ElseIf ((CICF == 0x0F))
       {
           CICM = SAMT /* \SAMT */
       }
       ElseIf ((CICF == 0x07))
       {
           CICM = SADM /* \SADM */
       }
       Else
       {
           CICM = 0x0F
       }

       Local1 |= (CICM << 0x0C)
       Local1 |= (Local5 << 0x10)
       Local1 |= 0x01
   }

thanks,


--
js
suse labs




[Index of Archives]     [Linux Kernel Development]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux