Re: [Qemu-devel] [PATCH v2 0/2] acpi unit-test: rebuild aml files functionality

[Marcel Apfelbaum]

On Fri, 2013-12-27 at 00:01 +0200, Marcel Apfelbaum wrote:
> On Thu, 2013-12-26 at 20:19 +0200, Michael S. Tsirkin wrote:
> > On Thu, Dec 26, 2013 at 07:15:00PM +0200, Marcel Apfelbaum wrote:
> > > On Thu, 2013-12-26 at 18:51 +0200, Michael S. Tsirkin wrote:
> > > > On Thu, Dec 26, 2013 at 06:12:25PM +0200, Michael S. Tsirkin wrote:
> > > > > On Thu, Dec 26, 2013 at 05:58:09PM +0200, Marcel Apfelbaum wrote:
> > > > > > On Thu, 2013-12-26 at 17:41 +0200, Michael S. Tsirkin wrote:
> > > > > > > On Thu, Dec 26, 2013 at 04:09:58PM +0200, Marcel Apfelbaum wrote:
> > > > > > > > On Thu, 2013-12-26 at 15:53 +0200, Michael S. Tsirkin wrote:
> > > > > > > > > On Thu, Dec 26, 2013 at 03:24:28PM +0200, Marcel Apfelbaum 
> > > > > > > > > wrote:
> > > > > > > > > > On Thu, 2013-12-26 at 14:46 +0200, Michael S. Tsirkin wrote:
> > > > > > > > > > > On Thu, Dec 26, 2013 at 02:19:48PM +0200, Marcel 
> > > > > > > > > > > Apfelbaum wrote:
> > > > > > > > > > > > Acpi unit-test will fail every time the acpi tables 
> > > > > > > > > > > > change.
> > > > > > > > > > > > 
> > > > > > > > > > > > The series adds a script that rebuilds the expected aml 
> > > > > > > > > > > > files, so the test
> > > > > > > > > > > > will pass. It also validates the modifications.
> > > > > > > > > > > > 
> > > > > > > > > > > > The acpi unit test will rebuild the aml tables if 
> > > > > > > > > > > > TEST_ACPI_REBUILD_AML
> > > > > > > > > > > > environment variable is set.
> > > > > > > > > > > 
> > > > > > > > > > > OK I had to fix some trailing whitespace, otherwise
> > > > > > > > > > > looks good.
> > > > > > > > > > > I applied it all, pls check the pci branch.
> > > > > > > > > > Thanks!
> > > > > > > > > > I pulled the pci branch, but I did not see the patches yet. 
> > > > > > > > > > I used git://git.kernel.org/pub/scm/virt/kvm/mst/qemu.git,
> > > > > > > > > > pci branch.
> > > > > > > > > > 
> > > > > > > > > > Thanks,
> > > > > > > > > > Marcel
> > > > > > > > > 
> > > > > > > > > 
> > > > > > > > > Actually had to revert. Build fails with recent gcc:
> > > > > > > > Strange, I update the gcc to the Fedora's latest 4.8.2 and
> > > > > > > > it still doesn't happen.
> > > > > > > > Anyway, I use qemu_write_full as advised and resent.
> > > > > > > > 
> > > > > > > > Thanks,
> > > > > > > > Marcel
> > > > > > > 
> > > > > > > OK I applied this for now but I see an issue: iasl crashes below.
> > > > > > Thanks!
> > > > > > By the way, I resend the whole series, I hope you used it.
> > > > > 
> > > > > yes pushed, take a look
> > > > > 
> > > > > > > I think this happens when we run it on a file
> > > > > > > it does not understand.
> > > > > > > It's an iasl bug but would be nice to work around
> > > > > > > it in some way (and it would be nice to report to
> > > > > > > robert moore)
> > > > > > 
> > > > > > I cannot reproduce it, I'll try anyway. 
> > > > > > By the way, the test checks if iasl exits with error,
> > > > > > so it should handle this gracefully. (I hope!)
> > > > > > 
> > > > > > Regarding the IASL project, I'll see if I can open a bug,
> > > > > 
> > > > > no, just send mail.
> > > > > 
> > > > > > (if I can reproduce it, otherwise I risk they'll ask some
> > > > > > questions I cannot answer...)
> > > > > > 
> > > > > > Thanks,
> > > > > > Marcel
> > > > > 
> > > > > This is on fedora 19, 32 bit.
> > > > 
> > > > I see that the issue is with HPET table.
> > > > 
> > > > iasl -e DSDT -e SSDT -d HPET
> > > > 
> > > > crashes for me
> > > > 
> > > > iasl -d HPET
> > > > 
> > > > does not crash
> > > > 
> > > > I think that only DSDT and SSDT need -e.
> > > > 
> > > > Just skip -e for anything that is not an SSDT
> > > > and not a DSDT.
> > > Thanks for the tip. I'll send a fix soon.
> > > Marcel
> > > 
> > > > 
> > > 
> > > 
> > 
> > also I see you still do
> > 
> >            aml_file = g_strdup_printf("%s/%s/%.4s", data_dir,
> > data->machine,
> >                                        (gchar *)&sdt->header.signature);
> > 
> > 
> > this will not work if we have > 1 SSDT.
> As I stated in a prev thread:
> From what I understand there is only one ssdt table named SSDT,
> all others are named differently: "APIC" "HPET".
> I think multiple SSDTS have different signatures, even in my laptop
> they are called: SSDT1, SSDT2, SSDT3,...
I disassembled the SSDTx tables in my laptop and I saw that actually
they have the same "SSDT" signature.
I still believe what  I wrote before:
  1. There is only one ssdt table named SSDT in Qemu, all the others
     are named differently: "APIC" "HPET".
  2. Even if the SPEC allows to use the same signature, qemu we can use
     different signatures for ssdt tables (it is allowed). Otherwise it
     will complicate the code unnecessary.

Thanks,
 Marcel

 
> 
> 
> 
>