Re: [PATCH v3 1/8] arm64/sysreg: Add definitions for immediate versions of MSR ALLINT

From: "Liao, Chang" <liaochang1@huawei.com>
To: Mark Brown <broonie@kernel.org>, Mark Rutland <mark.rutland@arm.com>
Cc: <catalin.marinas@arm.com>, <will@kernel.org>, <maz@kernel.org>,
	<oliver.upton@linux.dev>, <james.morse@arm.com>,
	<suzuki.poulose@arm.com>, <yuzenghui@huawei.com>,
	<tglx@linutronix.de>, <ardb@kernel.org>,
	<anshuman.khandual@arm.com>, <miguel.luis@oracle.com>,
	<joey.gouly@arm.com>, <ryan.roberts@arm.com>,
	<jeremy.linton@arm.com>, <ericchancf@google.com>,
	<kristina.martsenko@arm.com>, <robh@kernel.org>,
	<scott@os.amperecomputing.com>, <songshuaishuai@tinylab.org>,
	<shijie@os.amperecomputing.com>, <akpm@linux-foundation.org>,
	<bhe@redhat.com>, <horms@kernel.org>, <mhiramat@kernel.org>,
	<rmk+kernel@armlinux.org.uk>, <shahuang@redhat.com>,
	<takakura@valinux.co.jp>, <dianders@chromium.org>,
	<swboyd@chromium.org>, <sumit.garg@linaro.org>,
	<frederic@kernel.org>, <reijiw@google.com>,
	<akihiko.odaki@daynix.com>, <ruanjinjie@huawei.com>,
	<linux-arm-kernel@lists.infradead.org>,
	<linux-kernel@vger.kernel.org>, <kvmarm@lists.linux.dev>
Subject: Re: [PATCH v3 1/8] arm64/sysreg: Add definitions for immediate versions of MSR ALLINT
Date: Tue, 7 May 2024 15:41:08 +0800	[thread overview]
Message-ID: <cde4d448-dc9d-eaad-4a2d-a6d34bda4449@huawei.com> (raw)
In-Reply-To: <Zjjz-tzLRC2nH51A@finisterre.sirena.org.uk>

Hi, Mark.

在 2024/5/6 23:15, Mark Brown 写道:
> On Fri, May 03, 2024 at 05:00:49PM +0100, Mark Rutland wrote:
>> On Mon, Apr 15, 2024 at 06:47:51AM +0000, Liao Chang wrote:
> 
>> +#define PSTATE_ALLINT                  pstate_field(1, 0)
> 
>> +#define set_pstate_allint(x)           asm volatile(SET_PSTATE_ALLINT(x))
> 
> Hrm, those helpers are not ideally discoverable, partly due to the
> system register description for ALLINT not providing any references to
> this being a general scheme (which is fixable there) and partly due to

Based on the Arm ISA reference manual, the instruction accessing the ALLINT
field of PSTATE uses the following encoding:
                    op0  op1   CRn    CRm    op2
MSR ALLINT, #<imm>  0b00 0b001 0b0100 0b000x 0b000

In this encoding, the 'x' represents the LSB of #<imm>, op1 is fixed as 0b001
and op2 is fixed as 0b000. With this understanding, those helpers seem like a
good approach for accessing the PSTATE.ALLINT field. I've aslo confirmed that
the binary encoding generated by those helpers is same with the encoding of v3.

> the use of __emit_inst() with a numeric literal - we should probably add
> a comment next to the __emit_inst() saying what instruction we are
> emitting
Arm Architecture Reference Manual for A-profile outlines two variants for
MSR instructions used to modify PSTATE fields direclty using immediate. The
major difference between these variants lies in the CRm field encoding:

- 4 bit immediate, examples include "MSR DAIFSET,#Imm4" and "MSR DAIFCLR,#Imm4".
  The CRm field in this variant uses the least 4 bits of immediate.
- 1 bit immediate, currently, only "MSR ALLINT,#Imm1" uses this variant.
  The CRm field uses only the least 1 bit of immediate.

The current implementation of the macro SET_PSTATE() defaults to the 1 bit
immediate variant (!!x << PSTATE_Imm_shift). Currently, this macro is used
to generate instructions accessing PAN, UAO, SSBS, TCO and DIT which require
1 bit immediate variant, hence I would say it also work for ALLINT as well.

Thanks.

-- 
BR
Liao, Chang

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel