On Wed, Sep 09, 2020 at 09:15:22AM +0200, Wolfgang Wallner wrote:
-----"Andy Shevchenko" andy.shevchenko@gmail.com schrieb: -----
Betreff: Re: [PATCH v1] cmd: acpi: Print revisions in hex format On Tue, Sep 08, 2020 at 05:32:08PM +0200, Wolfgang Wallner wrote:
-----"Andy Shevchenko" andy.shevchenko@gmail.com schrieb: -----
On Tue, Sep 8, 2020 at 5:58 PM Wolfgang Wallner wolfgang.wallner@br-automation.com wrote:
-----"Andy Shevchenko" andriy.shevchenko@linux.intel.com schrieb: -----
...
Related to "acpi list": During my recent ACPI debugging I found it very useful to have the checksum printed for each table with "acpi list". Would there be interest to have that upstream? If so I would send a patch.
Can you elaborate what was the problem that checksum helped?
Sure. I saw two strange things with the ACPI checksums:
- The DSDT length included uninitialized bytes from alignment. This is
described in the following link:
https://lists.denx.de/pipermail/u-boot/2020-September/425378.html
This was the actual bug I was looking for.
- acpi_create_spcr() is missing a memset(). The other acpi_create_xxxx()
functions perform a memset on their structure, acpi_create_spcr() does not and as a result the contents of this table are party uninitialized.
I plan to send a patch for both of them.
I'm not sure I understood how checksum pointed to uninitialized data?
After adding the checksums to "acpi list" I realized that the checksums for DSDT and SPCR where different after every reset. Looking at the code each turned out to be somehow related to uninitialized memory.
I see. I would rather to have it as a separate (sub-)command. But I leave this to Simon.