Re: [PATCH v5 5/6] kprobes: Use text_alloc() and text_free()

From: Mike Rapoport <rppt@kernel.org>
To: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Cc: Ingo Molnar <mingo@kernel.org>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org,
	Andi Kleen <ak@linux.intel.com>,
	Masami Hiramatsu <mhiramat@kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	"Naveen N. Rao" <naveen.n.rao@linux.ibm.com>,
	Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>,
	"David S. Miller" <davem@davemloft.net>,
	Jessica Yu <jeyu@kernel.org>, Ard Biesheuvel <ardb@kernel.org>
Subject: Re: [PATCH v5 5/6] kprobes: Use text_alloc() and text_free()
Date: Wed, 19 Aug 2020 09:47:18 +0300	[thread overview]
Message-ID: <20200819064718.GR752365@kernel.org> (raw)
In-Reply-To: <20200818163033.GF137138@linux.intel.com>

On Tue, Aug 18, 2020 at 07:30:33PM +0300, Jarkko Sakkinen wrote:
> On Tue, Aug 18, 2020 at 02:51:41PM +0300, Mike Rapoport wrote:
> > On Tue, Aug 18, 2020 at 08:30:29AM +0300, Jarkko Sakkinen wrote:
> > > On Sun, Jul 26, 2020 at 11:14:08AM +0300, Mike Rapoport wrote:
> > > > > 
> > > > > I'm not still sure that I fully understand this feedback as I don't see
> > > > > any inherent and obvious difference to the v4. In that version fallbacks
> > > > > are to module_alloc() and module_memfree() and text_alloc() and
> > > > > text_memfree() can be overridden by arch.
> > > > 
> > > > The major difference between your v4 and my suggestion is that I'm not
> > > > trying to impose a single ARCH_HAS_TEXT_ALLOC as an alternative to
> > > > MODULES but rather to use per subsystem config option, e.g.
> > > > HAVE_KPROBES_TEXT_ALLOC.
> > > > 
> > > > Another thing, which might be worth doing regardless of the outcome of
> > > > this discussion is to rename alloc_insn_pages() to text_alloc_kprobes()
> > > > because the former is way too generic and does not emphasize that the 
> > > > instruction page is actually used by kprobes only.
> > > 
> > > What if we in kernel/kprobes.c just:
> > > 
> > > #ifndef CONFIG_ARCH_HAS_TEXT_ALLOC
> > 
> > I don't think that CONFIG_ARCH_HAS_TEXT_ALLOC will work for all
> > architectures.
> > 
> > If an architecture has different restrictions for allocation of text
> > for, say, modules, kprobes, bfp, it won't be able to use a single
> > ARCH_HAS_TEXT_ALLOC. Which means that this architecture is stuck with
> > dependency of kprobes on MODULES until the next rework.
> 
> I was thinking to name it as CONFIG_HAS_KPROBES_ALLOC_PAGE, alogn the
> lines described below, so it is merely a glitch in my example.

IMHO, it would be better to emphasize that this is text page. I liked
Mark's idea [1] to have text_alloc_<subsys>() naming for the allocation
functions. The Kconfig options then would become
HAVE_TEXT_ALLOC_<SUBSYS>.

[1] https://lore.kernel.org/linux-riscv/20200714133314.GA67386@C02TD0UTHF1T.local/

> > > void __weak *alloc_insn_page(void)
> > > {
> > > 	return module_alloc(PAGE_SIZE);
> > > }
> > > 
> > > void __weak free_insn_page(void *page)
> > > {
> > > 	module_memfree(page);
> > > }
> > > #endif
> > > 
> > > In Kconfig (as in v5):
> > > 
> > > config KPROBES
> > > 	bool "Kprobes"
> > > 	depends on MODULES || ARCH_HAS_TEXT_ALLOC
> > > 
> > > I checked architectures that override alloc_insn_page(). With the
> > > exception of x86, they do not call module_alloc().
> > > 
> > > If no rename was done, then with this approach a more consistent.
> > > config flag name would be CONFIG_ARCH_HAS_ALLOC_INSN_PAGE.
> > > 
> > > I'd call the function just as kprobes_alloc_page(). Then the
> > > config flag would become CONFIG_HAS_KPROBES_ALLOC_PAGE.
> > > 
> > > > -- 
> > > > Sincerely yours,
> > > > Mike.
> > > 
> > > Thanks for the feedback!
> > > 
> > > /Jarkko
> > 
> 
> > -- 
> > Sincerely yours,
> > Mike.
> 
> BR,
> /Jarkko

-- 
Sincerely yours,
Mike.