Patchwork [20/20,v3] sparc/mm/fault_64.c: Port OOM changes to do_sparc64_fault

login
register
mail settings
Submitter Kautuk Consul
Date March 26, 2012, 3:33 p.m.
Message ID <1332775997-2197-1-git-send-email-consul.kautuk@gmail.com>
Download mbox | patch
Permalink /patch/148770/
State Not Applicable
Delegated to: David Miller
Headers show

Comments

Kautuk Consul - March 26, 2012, 3:33 p.m.
Commit d065bd810b6deb67d4897a14bfe21f8eb526ba99
(mm: retry page fault when blocking on disk transfer) and
commit 37b23e0525d393d48a7d59f870b3bc061a30ccdb
(x86,mm: make pagefault killable)

The above commits introduced changes into the x86 pagefault handler
for making the page fault handler retryable as well as killable.

These changes reduce the mmap_sem hold time, which is crucial
during OOM killer invocation.

Port these changes to 64-bit sparc.

Signed-off-by: Kautuk Consul <consul.kautuk@gmail.com>
---
 arch/sparc/mm/fault_64.c |   38 +++++++++++++++++++++++++++++++-------
 1 files changed, 31 insertions(+), 7 deletions(-)
Joe Perches - March 26, 2012, 4:31 p.m.
On Mon, 2012-03-26 at 11:33 -0400, Kautuk Consul wrote:
> diff --git a/arch/sparc/mm/fault_64.c b/arch/sparc/mm/fault_64.c
> index 504c062..8322962 100644
> --- a/arch/sparc/mm/fault_64.c
> +++ b/arch/sparc/mm/fault_64.c
> @@ -279,6 +279,7 @@ asmlinkage void __kprobes do_sparc64_fault(struct pt_regs *regs)
>  	unsigned int insn = 0;
>  	int si_code, fault_code, fault;
>  	unsigned long address, mm_rss;
> +	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
>  
>  	fault_code = get_thread_fault_code();
>  
[]
> @@ -423,7 +426,13 @@ good_area:
>  			goto bad_area;
>  	}
>  
> -	fault = handle_mm_fault(mm, vma, address, (fault_code & FAULT_CODE_WRITE) ? FAULT_FLAG_WRITE : 0);
> +	fault = handle_mm_fault(mm, vma, address,
> +							flags | ((fault_code & FAULT_CODE_WRITE) ?
> +									 FAULT_FLAG_WRITE : 0);

A bit redundant and very poor style argument alignment
Make sure your tab indentation is set to 8.

	fault = handle_mm_fault(mm, vma, address,
				flags | (fault_code & FAULT_CODE_WRITE));

though perhaps fault_code could be unsigned and
perhaps you could use the same style initializer
for fault_code as patch 19/20.


--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Kautuk Consul - March 26, 2012, 4:50 p.m.
Okay, I have sent both patches with tab-space as 8 characters and set the
multiline C statements after the preceding line's paranthesis.
--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
David Miller - March 26, 2012, 8:11 p.m.
From: Kautuk Consul <consul.kautuk@gmail.com>
Date: Mon, 26 Mar 2012 11:33:17 -0400

> +	fault = handle_mm_fault(mm, vma, address,
> +							flags | ((fault_code & FAULT_CODE_WRITE) ?
> +									 FAULT_FLAG_WRITE : 0);

This is so shockingly gross I'm at a loss for words.  Why are you
tabbing this crap all the way into outer-space?

Make the "flags | ..." line up with the first column after the
openning parenthesis of "handle_mm_fault(".

> +			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ,
> +						  1, regs, address);

Same problem.

> +			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN,
> +						  1, regs, address);

Same problem.

You're making the coding style worse each iteration, not better.
--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/arch/sparc/mm/fault_64.c b/arch/sparc/mm/fault_64.c
index 504c062..8322962 100644
--- a/arch/sparc/mm/fault_64.c
+++ b/arch/sparc/mm/fault_64.c
@@ -279,6 +279,7 @@  asmlinkage void __kprobes do_sparc64_fault(struct pt_regs *regs)
 	unsigned int insn = 0;
 	int si_code, fault_code, fault;
 	unsigned long address, mm_rss;
+	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 
 	fault_code = get_thread_fault_code();
 
@@ -333,6 +334,8 @@  asmlinkage void __kprobes do_sparc64_fault(struct pt_regs *regs)
 			insn = get_fault_insn(regs, insn);
 			goto handle_kernel_fault;
 		}
+
+retry:
 		down_read(&mm->mmap_sem);
 	}
 
@@ -423,7 +426,13 @@  good_area:
 			goto bad_area;
 	}
 
-	fault = handle_mm_fault(mm, vma, address, (fault_code & FAULT_CODE_WRITE) ? FAULT_FLAG_WRITE : 0);
+	fault = handle_mm_fault(mm, vma, address,
+							flags | ((fault_code & FAULT_CODE_WRITE) ?
+									 FAULT_FLAG_WRITE : 0);
+
+	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
+		return;
+
 	if (unlikely(fault & VM_FAULT_ERROR)) {
 		if (fault & VM_FAULT_OOM)
 			goto out_of_memory;
@@ -431,12 +440,27 @@  good_area:
 			goto do_sigbus;
 		BUG();
 	}
-	if (fault & VM_FAULT_MAJOR) {
-		current->maj_flt++;
-		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address);
-	} else {
-		current->min_flt++;
-		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address);
+
+	if (flags & FAULT_FLAG_ALLOW_RETRY) {
+		if (fault & VM_FAULT_MAJOR) {
+			current->maj_flt++;
+			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ,
+						  1, regs, address);
+		} else {
+			current->min_flt++;
+			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN,
+						  1, regs, address);
+		}
+		if (fault & VM_FAULT_RETRY) {
+			flags &= ~FAULT_FLAG_ALLOW_RETRY;
+
+			/* No need to up_read(&mm->mmap_sem) as we would
+			 * have already released it in __lock_page_or_retry
+			 * in mm/filemap.c.
+			 */
+
+			goto retry;
+		}
 	}
 	up_read(&mm->mmap_sem);