Re: [PATCH v4] mm: Avoid unnecessary page fault retires on shared memory types

From: Christian Borntraeger <borntraeger@linux.ibm.com>
To: Peter Xu <peterx@redhat.com>
Cc: Heiko Carstens <hca@linux.ibm.com>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org,
	Will Deacon <will@kernel.org>, Matt Turner <mattst88@gmail.com>,
	linux-s390@vger.kernel.org,
	Andrew Morton <akpm@linux-foundation.org>,
	Brian Cain <bcain@quicinc.com>, Borislav Petkov <bp@alien8.de>,
	linux-alpha@vger.kernel.org, Alistair Popple <apopple@nvidia.com>,
	Jonas Bonn <jonas@southpole.se>,
	linux-arm-kernel@lists.infradead.org, linux-ia64@vger.kernel.org,
	Michael Ellerman <mpe@ellerman.id.au>,
	Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>,
	linux-snps-arc@lists.infradead.org,
	Vineet Gupta <vgupta@kernel.org>,
	Vasily Gorbik <gor@linux.ibm.com>,
	Vlastimil Babka <vbabka@suse.cz>,
	Ivan Kokshaysky <ink@jurassic.park.msu.ru>,
	Rich Felker <dalias@libc.org>,
	sparclinux@vger.kernel.org, Russell King <linux@armlinux.org.uk>,
	David Hildenbrand <david@redhat.com>,
	Benjamin Herrenschmidt <benh@kernel.crashing.org>,
	Nicholas Piggin <npiggin@gmail.com>,
	"James E . J . Bottomley" <James.Bottomley@hansenpartnership.com>,
	linux-xtensa@linux-xtensa.org, linux-sh@vger.kernel.org,
	Paul Walmsley <paul.walmsley@sifive.com>,
	linux-m68k@lists.linux-m68k.org, linuxppc-dev@lists.ozlabs.org,
	Richard Henderson <rth@twiddle.net>, Guo Ren <guoren@kernel.org>,
	linux-parisc@vger.kernel.org,
	Andrea Arcangeli <aarcange@redhat.com>,
	Helge Deller <deller@gmx.de>, Al Viro <viro@zeniv.linux.org.uk>,
	Albert Ou <aou@eecs.berkeley.edu>,
	linux-um@lists.infradead.org, "H . Peter Anvin" <hpa@zytor.com>,
	Janosch Frank <frankja@linux.ibm.com>,
	Sven Schnelle <svens@linux.ibm.com>,
	Anton Ivanov <anton.ivanov@cambridgegreys.com>,
	openrisc@lists.librecores.org,
	Thomas Bogendoerfer <tsbogend@alpha.franken.de>,
	linux-hexagon@vger.kernel.org, Andy Lutomirski <luto@kernel.org>,
	Stafford Horne <shorne@gmail.com>,
	linux-csky@vger.kernel.org, Thomas Gleixner <tglx@linutronix.de>,
	linux-mips@vger.kernel.org, Paul Mackerras <paulus@samba.org>,
	Alexander Gordeev <agordeev@linux.ibm.com>,
	Dinh Nguyen <dinguyen@kernel.org>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	"David S . Miller" <davem@davemloft.net>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Hugh Dickins <hughd@google.com>, Ingo Molnar <mingo@redhat.com>,
	Peter Zijlstra <peterz@infradead.org>,
	linux-riscv@lists.infradead.org,
	Max Filippov <jcmvbkbc@gmail.com>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Geert Uytterhoeven <geert@linux-m68k.org>,
	Johannes Berg <johannes@sipsolutions.net>,
	Chris Zankel <chris@zankel.net>, Michal Simek <monstr@monstr.eu>,
	x86@kernel.org, Yoshinori Sato <ysato@users.sourceforge.jp>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Richard Weinberger <richard@nod.at>,
	Ingo Molnar <mingo@kernel.org>
Subject: Re: [PATCH v4] mm: Avoid unnecessary page fault retires on shared memory types
Date: Mon, 30 May 2022 20:29:30 +0200	[thread overview]
Message-ID: <58b163b1-22ec-7634-dcd3-5a0a54adec45@linux.ibm.com> (raw)
In-Reply-To: <YpTqNKMTt8PoA41n@xz-m1.local>

Am 30.05.22 um 18:00 schrieb Peter Xu:
> On Mon, May 30, 2022 at 11:52:54AM -0400, Peter Xu wrote:
>> On Mon, May 30, 2022 at 11:35:10AM +0200, Christian Borntraeger wrote:
>>>
>>>
>>> Am 29.05.22 um 22:33 schrieb Heiko Carstens:
>>> [...]
>>>>
>>>> Guess the patch below on top of your patch is what we want.
>>>> Just for clarification: if gmap is not NULL then the process is a kvm
>>>> process. So, depending on the workload, this optimization makes sense.
>>>>
>>>> diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c
>>>> index 4608cc962ecf..e1d40ca341b7 100644
>>>> --- a/arch/s390/mm/fault.c
>>>> +++ b/arch/s390/mm/fault.c
>>>> @@ -436,12 +436,11 @@ static inline vm_fault_t do_exception(struct pt_regs *regs, int access)
>>>>    	/* The fault is fully completed (including releasing mmap lock) */
>>>>    	if (fault & VM_FAULT_COMPLETED) {
>>>> -		/*
>>>> -		 * Gmap will need the mmap lock again, so retake it.  TODO:
>>>> -		 * only conditionally take the lock when CONFIG_PGSTE set.
>>>> -		 */
>>>> -		mmap_read_lock(mm);
>>>> -		goto out_gmap;
>>>> +		if (gmap) {
>>>> +			mmap_read_lock(mm);
>>>> +			goto out_gmap;
>>>> +		}
>>>> +		goto out;
> 
> Hmm, right after I replied I found "goto out" could be problematic, since
> all s390 callers of do_exception() will assume it an error condition (side
> note: "goto out_gmap" contains one step to clear "fault" to 0).  I'll
> replace this with "return 0" instead if it looks good to both of you.
> 
> I'll wait for a confirmation before reposting.  Thanks,

Yes, that sounds good and thank you for double checking.