[[email protected]: Re: TLB flushing in 2.4.14]
Hey folks,
I'm just going through my backlog, and this patch looks reasonable to me. Linus, could you drop this into 2.5?
-ben
----- Forwarded message from Jack Steiner <[email protected]> -----
From: Jack Steiner <[email protected]>
Subject: Re: TLB flushing in 2.4.14
To: [email protected] (Benjamin LaHaise)
Date: Wed, 28 Nov 2001 14:01:14 -0600 (CST)
In-Reply-To: <[email protected]> from "Benjamin LaHaise" at Nov 27, 2001 01:25:19 PM
X-Mailer: ELM [version 2.5 PL2]
Quoted 9 lines · Jack Steiner
Quoted 3 lines
Do you see any problems with this patch??
(I also sent a copy of this to Andrea Arcangeli)
Yes. Your system will now happily corrupt memory if you run ps or top at the right time. I'd suggested adding an
mm->deadflag and using that instead, but Linus is the keeper of the vm code, so run it by him.
You are right, of course. I should have studied the code better.
Here is a much cleaner fix (ignore the previous patch completely). This keeps all the changes in just one spot - no changes to fs/exec.c.
Do you see any problems here...
include/asm-generic/tlb.h +5 -2
@@ -43,7 +43,7 @@ tlb->mm = mm;/* Use fast mode if there is only one user of this mm (this process) */ - tlb->nr = (atomic_read (&(mm)->mm_users) ==1 ) ? ~0UL :0UL ; + tlb->nr = (atomic_read (&(mm)->mm_users) <=1 ) ? ~0UL :0UL ;return tlb; }
@@ -70,6 +70,8 @@/* tlb_finish_mmu * Called at the end of the shootdown operation to free up any resources * that were required. The page talbe lock is still held at this point. +* Note that no TLB flushes are needed if there are no users of the mm +* context. */ static inline void tlb_finish_mmu (struct free_pte_ctx *ctx,unsigned long start,unsigned long end) {
@@ -77,7 +79,8 @@/* Handle the fast case first. */ if (ctx->nr == ~0UL ) { -flush_tlb_range (ctx->mm, start, end); +if (ctx->mm->mmap) +flush_tlb_range (ctx->mm, start, end);return ; } nr = ctx->nr;
Signature · 12 lines
Thanks Jack Steiner (651-683-5302) (vnet 233-5302) [email protected] ----- End forwarded message ----- -- Fish. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/