plat-qcom: Introduce diagnostic ring buffer for post-mortem trace capture

[zelvam95]

Summary

On Qualcomm production devices the UART is typically disabled, and when
present it is shared between the non-secure (Linux kernel) and secure
(OP-TEE) execution environments. Consequently, the UART cannot serve as
a dedicated console for OP-TEE, and trace output is unavailable for
post-mortem analysis after a crash or reset.

This series introduces a diagnostic ring buffer in IMEM in which OP-TEE
records its log output. Because IMEM contents are preserved across a
warm reset, the captured trace can be retrieved out of band by an
external reader for debug and post-mortem use.

Design

The buffer is a write-only circular buffer placed in a fixed region of
IMEM. A header at the start of the region records the version, a magic
value indicating buffer state, the buffer offset and size, a wrap
counter, and the current write head. An external reader uses the
header to walk the buffer and reconstruct the log content.

The implementation exposes two functions:

• qcom_diag_log_init() initializes the buffer header and the
  write-only circular log region in IMEM. Logging is suppressed when
  the SoC is in DLOAD (download) mode, as indicated by the TCSR
  boot-misc register.
• qcom_diag_log_puts() appends a string to the circular buffer. On
  overflow the buffer wraps in place and a saturating wrap counter
  records the number of overruns.

These are wired into OP-TEE's trace subsystem by overriding the
plat_trace_init() and plat_trace_ext_puts() weak hooks defined in
core, so every trace line emitted by OP-TEE core is mirrored into the
IMEM buffer.

The feature is gated by CFG_QCOM_DIAG_LOG, which defaults to the
value of CFG_TEE_CORE_DEBUG on each enabled target. When disabled,
the buffer is not initialized and the trace hooks are no-ops.

Patches in this series

1. plat-qcom: Add support for diagnostic ring buffer
   Introduce the diagnostic ring buffer in IMEM, including the
   qcom_diag_log_init() and qcom_diag_log_puts() interface.

2. plat-qcom: Wire diagnostic ring buffer into OP-TEE trace path
   Override plat_trace_init() and plat_trace_ext_puts() so that
   OP-TEE trace output is captured in the ring buffer.

3. plat-qcom: bobcat: Enable diagnostic ring buffer on ipq96xx and ipq52xx
   Add the buffer layout and per-target IMEM base/size for the bobcat
   architecture, and default CFG_QCOM_DIAG_LOG to
   CFG_TEE_CORE_DEBUG on both targets.

4. plat-qcom: hoya: Enable diagnostic ring buffer on kodiak and lemans
   Same as above for the hoya architecture and its kodiak / lemans
   targets.

Testing

• Built debug images for hoya/kodiak, hoya/lemans, bobcat/ipq96xx, and
  bobcat/ipq52xx with CFG_QCOM_DIAG_LOG=y and confirmed OP-TEE trace
  output is captured in the IMEM ring buffer and survives a warm
  reset.
• Built release images with CFG_QCOM_DIAG_LOG=n (the default for
  non-debug builds) and confirmed there is no behavioural change and
  no IMEM region is reserved.
• Verified that DLOAD mode (TCSR boot-misc register set to the
  download cookie) suppresses initialization of the buffer.

[github-actions]

FYI @ldts @b49020

[zelvam95]

Gentle Reminder to help review this and share any comments @ldts / @b49020

[b49020]

@zelvam95 please add CI build option for this DIAG driver. Similarly we need CI to build fuseprov drivers too, can you add the build config option for that too?

For both options, you can use lemans target.

[zelvam95]

@zelvam95 please add CI build option for this DIAG driver. Similarly we need CI to build fuseprov drivers too, can you add the build config option for that too?

For both options, you can use lemans target.

Have added the CI build option for both these drivers for lemans target. Let me know if there is any other feedback.

[zelvam95]

Gentle Reminder to share any other review comments/feedback @ldts / @b49020

[b49020]

Based on review already done for this patch-set here: qualcomm-linux/optee_os#7, feel free to add following for the series:

Reviewed-by: Sumit Garg <sumit.garg@oss.qualcomm.com>

[zelvam95]

Based on review already done for this patch-set here: qualcomm-linux/optee_os#7, feel free to add following for the series:

Reviewed-by: Sumit Garg <sumit.garg@oss.qualcomm.com>

Thanks Sumit, Have added the tags.

@jenswikl, Please let me know if you have any feedback/comments and if not, requesting to kindly help merge this.

[zelvam95]

@b49020 / @ldts , Have added one more FUSEPROV config clean up commit based on the feedback I received for DIAG Log^. Please help to review that patch as well & the latest commits and let me know if you have any comments.

@jenswikl, Have addressed the review comments as part of the latest patchset pushed. Let me know if this looks good & if you have any other feedback?

[ldts]

Reviewed-by: Jorge Ramirez-Ortiz <jorge.ramirez@oss.qualcomm.com>

[zelvam95]

Thanks @ldts. Have added the tag.
@b49020, Have addressed the fuseprov config enforcement. Let me know if you have any other feedback.

[b49020]

For the QFPROM commit:

Reviewed-by: Sumit Garg <sumit.garg@oss.qualcomm.com>

[zelvam95]

Thanks @b49020.
@jenswikl, Gentle Reminder - Let me know if you have any other specific feedback/review comments.

[jenswikl]

Looks good, I'll merge this when CI tests have passed.