From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:48257) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1gRwjd-0005ru-5q for qemu-devel@nongnu.org; Wed, 28 Nov 2018 05:06:22 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1gRwjb-0002iu-6g for qemu-devel@nongnu.org; Wed, 28 Nov 2018 05:06:21 -0500 Date: Wed, 28 Nov 2018 11:05:36 +0100 From: Samuel Ortiz Message-ID: <20181128100536.GC5677@caravaggio> References: <20181126162942.21258-1-sameo@linux.intel.com> <20181126162942.21258-5-sameo@linux.intel.com> <20181127162551.29608eeb@redhat.com> <20181127154218.GA5677@caravaggio> <20181127172749.75a036b4@redhat.com> <20181127221515-mutt-send-email-mst@kernel.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20181127221515-mutt-send-email-mst@kernel.org> Subject: Re: [Qemu-devel] [PATCH 4/8] hw: arm: Carry RSDP specific data through AcpiRsdpData List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: "Michael S. Tsirkin" Cc: Igor Mammedov , qemu-devel@nongnu.org, Peter Maydell , qemu-arm@nongnu.org, Shannon Zhao , Laurent Vivier , Richard Henderson , Paolo Bonzini , Ben Warren , Thomas Huth , Marcel Apfelbaum , Eduardo Habkost Hi Michael, On Tue, Nov 27, 2018 at 10:26:30PM -0500, Michael S. Tsirkin wrote: > On Tue, Nov 27, 2018 at 05:27:49PM +0100, Igor Mammedov wrote: > > On Tue, 27 Nov 2018 16:42:18 +0100 > > Samuel Ortiz wrote: > > > > > Hi Igor, > > > > > > On Tue, Nov 27, 2018 at 04:25:51PM +0100, Igor Mammedov wrote: > > > > On Mon, 26 Nov 2018 17:29:37 +0100 > > > > Samuel Ortiz wrote: > > > > > > > > > That will allow us to generalize the ARM build_rsdp() routine to support > > > > > both legacy RSDP (The current i386 implementation) and extended RSDP > > > > > (The ARM implementation). > > > > > > > > > > Signed-off-by: Samuel Ortiz > > > > > --- > > > > > include/hw/acpi/acpi-defs.h | 11 +++++++++++ > > > > > hw/arm/virt-acpi-build.c | 27 ++++++++++++++++++++++----- > > > > > 2 files changed, 33 insertions(+), 5 deletions(-) > > > > > > > > > > diff --git a/include/hw/acpi/acpi-defs.h b/include/hw/acpi/acpi-defs.h > > > > > index af8e023968..e7fd24c6c5 100644 > > > > > --- a/include/hw/acpi/acpi-defs.h > > > > > +++ b/include/hw/acpi/acpi-defs.h > > > > > @@ -53,6 +53,17 @@ struct AcpiRsdpDescriptor { /* Root System Descriptor Pointer */ > > > > > } QEMU_PACKED; > > > > > typedef struct AcpiRsdpDescriptor AcpiRsdpDescriptor; > > > > > > > > > > +typedef struct AcpiRsdpData { > > > > > + uint8_t oem_id[6]; /* OEM identification */ > > > > > + uint8_t revision; /* Must be 0 for 1.0, 2 for 2.0 */ > > > > > + > > > > > + unsigned *rsdt_tbl_offset; > > > > > + unsigned *xsdt_tbl_offset; > > > > > +} AcpiRsdpData; > > > > > + > > > > > > > > > +#define ACPI_RSDP_REV_1 0 > > > > > +#define ACPI_RSDP_REV_2 2 > > > > it's one time used spec defined values so just use values directly > > > > in place with a comment, so reader won't have to jump around code > > > > when comparing to spec. > > > It's also used in the ACPI tests fix patch. > > it's better to use in test it's own version (we just opencode them there) > > see fadt_fetch_facs_and_dsdt_ptrs()/sanitize_fadt_ptrs() > > same applies for length. > > that way if we break it in qemu's code test would catch the thing > > > > > Also the 0 for revision 1 is a little confusing, I feel the above > > > definition is clearer. > > that's confusion is in the spec, so we just mimic it, no need to add more on top > > To be more precise, there is a huge number of constants in ACPI > such that adding defines for them all would be a huge burden, I find that defining a set of well named constants is a lot less painful than maintaining code with at least the same amount of hard coded constants. That's a personal opinion, for sure. > and will not make it easy to check values against the > spec at all (case in point ACPI_RSDP_REV_2 is actually wrong, > 2 is version 3 and up). I may be misreading the spec, but I understand 0 is for ACPI 1.0 and 2 is for ACPI 2.0+. The latest spec is a little confusing with regard to this field, but when looking at the 2.0a ACPI spec for RSDP: "The ACPI version 1.0 revision number of this table is zero. The ACPI 2.0 value for this field is 2." > Thus the preferred style is to add a comment near the value > matching spec name verbatim, so one can copy it and > look it up in the spec. Sometimes one needs to reference > specific spec version. > > 0 /* Revision: ACPI version 1.0 */ > > and > > 1 /* Revision: ACPI 2.0 */ > > and > > 2 /* Revision: ACPI 3.0a */ > > For style consistency, if the value is used multiple times, we avoid > duplication by using an inline function and not a macro. Not entirely sure how this materializes. Do you mean that e.g. if I want to check for an RSDP revision I'd have to define inline functions of that kind: bool is_rsdp_revision_0(uint8_t *rsdp_table); bool is_rsdp_revision_2(uint8_t *rsdp_table); or do you have something else in mind? Cheers, Samuel.