Hi all,

Since kernel >= 6.2.15 I have to add "tpm_tis.interrupts=0" to successfully boot up.

Without "tpm_tis.interrupts=0" the system freezes right after taking over from the boot loader, no graphical output to diagnose the problem.
But: It seems to be a race. 1 out of 10 boot tries might be successful without any changes.

Hardware:
- Lenovo L590 laptop
- Secure boot is disabled
- Disk is encrypted (LUKS)
- TPM information below

The first graphical output on my system is normally the "enter the password" screen to unlock the disk encryption. Waiting for some time and then typing the password "into the dark" does not work.

I started a git bisect session on the 6.2 stable tree and this is the result:

32e3f11f2888b62a690596865598c8952916b6b3 is the first bad commit
commit 32e3f11f2888b62a690596865598c8952916b6b3
Author: Lino Sanfilippo <l.sanfilippo>
Date:   Thu Nov 24 14:55:29 2022 +0100

    tpm, tpm_tis: Claim locality before writing interrupt registers
    
    [ Upstream commit 15d7aa4e46eba87242a320f39773aa16faddadee ]
    
    In tpm_tis_probe_single_irq() interrupt registers TPM_INT_VECTOR,
    TPM_INT_STATUS and TPM_INT_ENABLE are modified to setup the interrupts.
    Currently these modifications are done without holding a locality thus they
    have no effect. Fix this by claiming the (default) locality before the
    registers are written.
    
    Since now tpm_tis_gen_interrupt() is called with the locality already
    claimed remove locality request and release from this function.
    
    Signed-off-by: Lino Sanfilippo <l.sanfilippo>
    Tested-by: Jarkko Sakkinen <jarkko>
    Reviewed-by: Jarkko Sakkinen <jarkko>
    Signed-off-by: Jarkko Sakkinen <jarkko>
    Stable-dep-of: 955df4f87760 ("tpm, tpm_tis: Claim locality when interrupts are reenabled on resume")
    Signed-off-by: Sasha Levin <sashal>

The same commit is in the 6.3 stable tree as cc1b06934237ed3fb2b9f75a577fd102f2dba471.

I went on and tried to revert that commit on top of 6.3.8.
The simple git operation failed due to conflicts.
The manual revert (which helps to solve my problem but might be faulty) is:

diff --git a/drivers/char/tpm/tpm_tis_core.c b/drivers/char/tpm/tpm_tis_core.c
index e05d2b227de3..74d96449ff46 100644
--- a/drivers/char/tpm/tpm_tis_core.c
+++ b/drivers/char/tpm/tpm_tis_core.c
@@ -1153,23 +1153,21 @@ int tpm_tis_core_init(struct device *dev, struct tpm_tis_data *priv, int irq,
                        goto out_err;
                }
 
-               if (irq)
+               if (irq) {
                        tpm_tis_probe_irq_single(chip, intmask, IRQF_SHARED,
                                                 irq);
-               else
-                       tpm_tis_probe_irq(chip, intmask);
-
-               if (chip->flags & TPM_CHIP_FLAG_IRQ) {
-                       priv->int_mask = intmask;
-               } else {
-                       dev_err(&chip->dev, FW_BUG
+                       if (!(chip->flags & TPM_CHIP_FLAG_IRQ)) {
+                               dev_err(&chip->dev, FW_BUG
                                        "TPM interrupt not working, polling instead\n");
 
-                       rc = tpm_tis_request_locality(chip, 0);
-                       if (rc < 0)
-                               goto out_err;
-                       disable_interrupts(chip);
-                       tpm_tis_relinquish_locality(chip, 0);
+                               rc = tpm_tis_request_locality(chip, 0);
+                               if (rc < 0)
+                                       goto out_err;
+                               disable_interrupts(chip);
+                               tpm_tis_relinquish_locality(chip, 0);
+                       }
+               } else {
+                       tpm_tis_probe_irq(chip, intmask);
                }
        }
 

When booting up with 6.2.14 I have the following TPM stuff in my boot log:

[    0.013514] ACPI: TPM2 0x000000007DC66000 000034 (v04 LENOVO TP-R0Z   00001200 PTEC 00000002)
[    0.013637] ACPI: Reserving TPM2 table memory at [mem 0x7dc66000-0x7dc66033]
[    1.197242] tpm_tis STM7364:00: 2.0 TPM (device-id 0x0, rev-id 78)
[    1.198978] tpm tpm0: [Firmware Bug]: TPM interrupt not working, polling instead


With the revert (from above) applied the "TPM interrupt not working" line is gone.

My temporary workaround is adding "tpm_tis.interrupts=0" for kernels > 6.2.14.
That makes my system booting up reliably.

I think this is a regression and should be reported upstream, no?

Reproducible: Always

Steps to Reproduce:
1. Reboot
2. Select a kernel >= 6.2.15 in boot menu
3. Make sure "tpm_tis.interrupts=0" is not part of the kernel cmdline options

Actual Results:  
The system does not boot up, it seems to hang after uncompressing the kernel/initrd. No graphical output.

Expected Results:  
Successful system boot


Note: This bug report has been updated. v1 was wrong, "nomodeset" is completely unrelated.