Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Michel Lespinasse <michel@lespinasse.org>
To:     Linux-MM <linux-mm@kvack.org>
Cc:     Laurent Dufour <ldufour@linux.ibm.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Michal Hocko <mhocko@suse.com>,
        Matthew Wilcox <willy@infradead.org>,
        Rik van Riel <riel@surriel.com>,
        Paul McKenney <paulmck@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Suren Baghdasaryan <surenb@google.com>,
        Joel Fernandes <joelaf@google.com>,
        Rom Lemarchand <romlem@google.com>,
        Linux-Kernel <linux-kernel@vger.kernel.org>,
        Michel Lespinasse <michel@lespinasse.org>
Subject: [RFC PATCH 33/37] mm: enable speculative fault handling only for multithreaded user space
Date:   Tue,  6 Apr 2021 18:44:58 -0700
Message-Id: <20210407014502.24091-34-michel@lespinasse.org>
In-Reply-To: <20210407014502.24091-1-michel@lespinasse.org>
References: <20210407014502.24091-1-michel@lespinasse.org>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

Performance tuning: single threaded userspace does not benefit from
speculative page faults, so we turn them off to avoid any related
(small) extra overheads.

Signed-off-by: Michel Lespinasse <michel@lespinasse.org>
---
 arch/x86/mm/fault.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 48b86911a6df..b1a07ca82d59 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -1318,6 +1318,10 @@ void do_user_addr_fault(struct pt_regs *regs,
 	}
 #endif
 
+	/* Only try spf for multithreaded user space faults. */
+	if (!(flags & FAULT_FLAG_USER) || atomic_read(&mm->mm_users) == 1)
+		goto no_spf;
+
 	count_vm_event(SPF_ATTEMPT);
 	seq = mmap_seq_read_start(mm);
 	if (seq & 1)
@@ -1351,6 +1355,7 @@ void do_user_addr_fault(struct pt_regs *regs,
 
 spf_abort:
 	count_vm_event(SPF_ABORT);
+no_spf:
 
 	/*
 	 * Kernel-mode access to the user address space should only occur
-- 
2.20.1