Re: [PATCH 6/8] SFI: add ACPI extensions

[Len Brown <lenb@kernel.org>]

On Tue, 23 Jun 2009, Andi Kleen wrote:
> Len Brown <lenb@kernel.org> writes:

> > diff --git a/drivers/acpi/tables.c b/drivers/acpi/tables.c
> > @@ -277,6 +277,9 @@ int __init acpi_table_parse(char *id, acpi_table_handler handler)

> > +	if (acpi_disabled)
> > +		return 1;

> This seems like a weird place to hook this in. Shouldn't that be somewhere else, more
> high level?

acpi_table_parse() is actually a high-level API available to drivers.

Today, drivers tend to test acpi_disabled on their own,
which may be too high level...

The catalyst for this change, IIR, was the PCI code calling
and not checking the return value:

--- a/arch/x86/pci/mmconfig-shared.c
...
@@ -606,7 +607,8 @@ static void __init __pci_mmcfg_init(int early)
...
-               acpi_table_parse(ACPI_SIG_MCFG, pci_parse_mcfg);
+               if (acpi_table_parse(ACPI_SIG_MCFG, pci_parse_mcfg))
+                       sfi_acpi_table_parse(ACPI_SIG_MCFG, NULL, NULL, 0, pci_parse_mcfg);
...


> > +	num_entries = (xsdt->header.length - sizeof(struct acpi_table_header) /
> > +		sizeof(u64));
> > +
> > +	pr_debug(PREFIX "XSDT has %d entries\n", num_entries);
> > +
> > +	for (i = 0; i < num_entries; i++)
> > +		sfi_tb_install_table(xsdt->table_offset_entry[i], SFI_ACPI_TABLE);
> 
> Shouldn't this have some more sanity checking, e.g. for overflows?

good observation. Although we did already check the signature and 
checksum, I don't see that we yet have a sanity check either here
or in sfi_tb_install_table() itself.

thanks,
-Len