Hello Robert, Jordan,

the S3 and S4 packages prepared by OVMF / edk2 commit [1] (svn rev
14003) look like

      Name (\_S3, Package (0x01)
      {
          0x00000001
      })
      Name (\_S4, Package (0x01)
      {
          0x00000002
      })

They are refused by the following ACPICA code section (please excuse the
wide listing):

repo: git://github.com/otcshare/acpica.git
file: source/components/hardware/hwxface.c

4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 635)     /*
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 636)      * The package must have at least two elements. NOTE (March 2005): This
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 637)      * goes against the current ACPI spec which defines this object as a
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 638)      * package with one encoded DWORD element. However, existing practice
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 639)      * by BIOS vendors seems to be to have 2 or more elements, at least
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 640)      * one per sleep type (A/B).
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 641)      */
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 642)     else if (Info->ReturnObject->Package.Count < 2)
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 643)     {
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 644)         ACPI_ERROR ((AE_INFO,
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 645)             "Sleep State return package does not have at least two elements"));
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 646)         Status = AE_AML_NO_OPERAND;
4fa895bc (Robert Moore 2008-11-19 13:15:10 -0800 647)     }

I ran into the problem with the 20121217.16 Fedora 18 nightly build:

[    0.000000] Linux version 3.6.10-4.fc18.x86_64 (mockbuild@) (gcc version 4.7.2 20121109 (Red Hat 4.7.2-8) (GCC) ) #1 SMP Tue Dec 11 18:01:27 UTC 2012

[    0.064483] ACPI: Interpreter enabled
[    0.064483] ACPI: (supports S0ACPI Error: Sleep State return package does not have at least two elements (20120711/hwxface-511)
[    0.064483] ACPI Exception: AE_AML_NO_OPERAND, While evaluating SleepState [\_S3_], bad Sleep object ffff880119a33c18 type Package (20120711/hwxface-543)
[    0.064483] ACPI Error: Sleep State return package does not have at least two elements (20120711/hwxface-511)
[    0.064483] ACPI Exception: AE_AML_NO_OPERAND, While evaluating SleepState [\_S4_], bad Sleep object ffff880119a33cf0 type Package (20120711/hwxface-543)
[    0.064483]  S5)


Should I attempt a patch for ACPICA (see the attachment), or is it not
worth the hassle and I should just fixup the AML generation in OVMF
(split the DWord into Bytes)?

(Please keep me CC'd, I'm not subscribed to <devel(a)acpica.org>.)

Thanks
Laszlo

[1] http://tianocore.git.sourceforge.net/git/gitweb.cgi?p=tianocore/edk2;a=commitdiff;h=14430c55c8d0e9a8487c2c74155e63299632ef5e