[PATCH 1/1 V2] Add Thread Support for the Context ID Register of ARM v6 & v7 Architectures

Message: 6
Date: Thu, 14 Jul 2011 11:33:10 +0200
From: Wolfgang BETZ <redacted>
To: "linux-arm-kernel at lists.infradead.org"
	[off-list ref], 	"linux at arm.linux.org.uk"
	[off-list ref]
Cc: "Rudolf.Dienstbeck at Lauterbach.com"
	[off-list ref],	Wolfgang BETZ
	[off-list ref],	Linus WALLEIJ [off-list ref],
	Srinidhi KASAGAR [off-list ref],
	"khaled.jmal at lauterbach.com" [off-list ref],
	"marco.ferrario at lauterbach.it" [off-list ref],
	"will.deacon at arm.com" [off-list ref],
	"maurizio.menegotto at lauterbach.it" [off-list ref],
	David SIORPAES [off-list ref],	Giuseppe DESOLI
	[off-list ref]
Subject: [PATCH 1/1 V2] Add Thread Support for the Context ID Register
	of	ARM v6 & v7 Architectures
Message-ID:
	[ref]

Content-Type: text/plain; charset="iso-8859-1"

From: Wolfgang Betz <redacted>

Hi Wolfgang,

sorry for replying through the digest, but I have comments on this one.

The aim of this patch is to enable thread support in the context ID register
(CONTEXTIDR) as it comes with ARM architectures v6 & v7.

[ ... ]

+/*
+ * Set context ID for task and mm
+ */
+static inline
+void set_context_id(struct task_struct *tsk, struct mm_struct *mm)
+{
+	unsigned int ctxid = (unsigned int)calc_context_id(tsk, mm);
+
+	/* set the new ContextID */
+	asm("mcr	p15, 0, %0, c13, c0, 1\n" : : "r" (ctxid));
+	isb();
+}

While I'm not qualified to comment on the technical correctness of the 
patch, I've got a few questions about the way this change is done:

Specifically:

1. about the above, (unsigned int)calc_context_id(), the cast and/or
    the data type isn't required, the asm doesn't care about types.

Generally:

2. the patch changes the signature of cpu_switch_mm() but not all the
    callers (doesn't touch the kexec / reset / suspend paths), why ?

3. what does this patch do to non-v6/v7 architectures ? Are they all
    _guaranteed_ never to use the second arg ?

    what about future architectures ? Is it a good idea to force the
    2nd arg to an integer where right now it's a pointer to a data
    structure, possibly far more flexible at passing "any info" ?

4. with that patch in, calling cpu_switch_mm() requires availability
    of "current" because of the proxy-wrapping, call calc_context_id()
    before which uses "current".
    There are usecases (kexec, reset, hibernate) where code that wants
    to cpu_switch_mm doesn't have "current" because of a stack switch;
    how are these codepaths supposed to call it ? Is it ok to just
    pass NULL / garbage ?

FrankH.