[34-longterm 190/260] De-pessimize rds_page_copy_user

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
From: Paul Gortmaker
Date: Sunday, January 2, 2011 - 12:18 am

From: Linus Torvalds <torvalds@linux-foundation.org>

commit 799c10559d60f159ab2232203f222f18fa3c4a5f upstream.

Don't try to "optimize" rds_page_copy_user() by using kmap_atomic() and
the unsafe atomic user mode accessor functions.  It's actually slower
than the straightforward code on any reasonable modern CPU.

Back when the code was written (although probably not by the time it was
actually merged, though), 32-bit x86 may have been the dominant
architecture.  And there kmap_atomic() can be a lot faster than kmap()
(unless you have very good locality, in which case the virtual address
caching by kmap() can overcome all the downsides).

But these days, x86-64 may not be more populous, but it's getting there
(and if you care about performance, it's definitely already there -
you'd have upgraded your CPU's already in the last few years).  And on
x86-64, the non-kmap_atomic() version is faster, simply because the code
is simpler and doesn't have the "re-try page fault" case.

People with old hardware are not likely to care about RDS anyway, and
the optimization for the 32-bit case is simply buggy, since it doesn't
verify the user addresses properly.

Reported-by: Dan Rosenberg <drosenberg@vsecurity.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
---
 net/rds/page.c |   27 +++++++--------------------
 1 files changed, 7 insertions(+), 20 deletions(-)

diff --git a/net/rds/page.c b/net/rds/page.c
index 595a952..1dfbfea 100644
--- a/net/rds/page.c
+++ b/net/rds/page.c
@@ -57,30 +57,17 @@ int rds_page_copy_user(struct page *page, unsigned long offset,
 	unsigned long ret;
 	void *addr;
 
-	if (to_user)
+	addr = kmap(page);
+	if (to_user) {
 		rds_stats_add(s_copy_to_user, bytes);
-	else
+		ret = copy_to_user(ptr, addr + offset, bytes);
+	} else {
 		rds_stats_add(s_copy_from_user, bytes);
-
-	addr = kmap_atomic(page, KM_USER0);
-	if (to_user)
-		ret = __copy_to_user_inatomic(ptr, addr + offset, bytes);
-	else
-		ret = __copy_from_user_inatomic(addr + offset, ptr, bytes);
-	kunmap_atomic(addr, KM_USER0);
-
-	if (ret) {
-		addr = kmap(page);
-		if (to_user)
-			ret = copy_to_user(ptr, addr + offset, bytes);
-		else
-			ret = copy_from_user(addr + offset, ptr, bytes);
-		kunmap(page);
-		if (ret)
-			return -EFAULT;
+		ret = copy_from_user(addr + offset, ptr, bytes);
 	}
+	kunmap(page);
 
-	return 0;
+	return ret ? -EFAULT : 0;
 }
 EXPORT_SYMBOL_GPL(rds_page_copy_user);
 
-- 
1.7.3.3

--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
[34-longterm 000/260] v2.6.34.8 longterm review, Paul Gortmaker, (Sun Jan 2, 12:14 am)
[34-longterm 003/260] ath5k: drop warning on jumbo frames, Paul Gortmaker, (Sun Jan 2, 12:14 am)
[34-longterm 019/260] ext4: Show journal_checksum option, Paul Gortmaker, (Sun Jan 2, 12:15 am)
[34-longterm 025/260] ext4: Fix compat EXT4_IOC_ADD_GROUP, Paul Gortmaker, (Sun Jan 2, 12:15 am)
[34-longterm 028/260] ext4: fix freeze deadlock under IO, Paul Gortmaker, (Sun Jan 2, 12:15 am)
[34-longterm 030/260] xen: handle events as edge-triggered, Paul Gortmaker, (Sun Jan 2, 12:15 am)
[34-longterm 045/260] USB: ehci-ppc-of: problems in unwind, Paul Gortmaker, (Sun Jan 2, 12:15 am)
[34-longterm 047/260] USB: CP210x Add new device ID, Paul Gortmaker, (Sun Jan 2, 12:15 am)
[34-longterm 065/260] irda: off by one, Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 089/260] sched: Optimize task_rq_lock(), Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 090/260] sched: Fix nr_uninterruptible count, Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 093/260] sched: Fix select_idle_sibling(), Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 098/260] arm: fix really nasty sigreturn bug, Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 106/260] drm/i915: Prevent double dpms on, Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 110/260] gro: fix different skb headrooms, Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 111/260] gro: Re-fix different skb headrooms, Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 115/260] tcp: fix three tcp sysctls tuning, Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 118/260] rds: fix a leak of kernel memory, Paul Gortmaker, (Sun Jan 2, 12:16 am)
[34-longterm 126/260] Staging: vt6655: fix buffer overflow, Paul Gortmaker, (Sun Jan 2, 12:17 am)
[34-longterm 134/260] percpu: fix pcpu_last_unit_cpu, Paul Gortmaker, (Sun Jan 2, 12:17 am)
[34-longterm 136/260] inotify: send IN_UNMOUNT events, Paul Gortmaker, (Sun Jan 2, 12:17 am)
[34-longterm 139/260] fix siglock, Paul Gortmaker, (Sun Jan 2, 12:17 am)
[34-longterm 145/260] AT91: change dma resource index, Paul Gortmaker, (Sun Jan 2, 12:17 am)
[34-longterm 154/260] inotify: fix inotify oneshot support, Paul Gortmaker, (Sun Jan 2, 12:17 am)
[34-longterm 158/260] alpha: Fix printk format errors, Paul Gortmaker, (Sun Jan 2, 12:17 am)
[34-longterm 188/260] atl1: fix resume, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 190/260] De-pessimize rds_page_copy_user, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 192/260] xfrm4: strip ECN bits from tos field, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 193/260] tcp: Fix &gt;4GB writes on 64-bit., Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 199/260] tcp: Fix race in tcp_poll, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 200/260] netxen: dont set skb-&gt;truesize, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 203/260] skge: add quirk to limit DMA, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 208/260] b44: fix carrier detection on bind, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 224/260] bluetooth: Fix missing NULL check, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 236/260] KVM: x86: Fix SVM VMCB reset, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 240/260] p54usb: fix off-by-one on !CONFIG_PM, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 241/260] p54usb: add five more USBIDs, Paul Gortmaker, (Sun Jan 2, 12:18 am)
[34-longterm 256/260] libsas: fix NCQ mixing with non-NCQ, Paul Gortmaker, (Sun Jan 2, 12:19 am)
[34-longterm 257/260] gdth: integer overflow in ioctl, Paul Gortmaker, (Sun Jan 2, 12:19 am)
[34-longterm 258/260] Fix race when removing SCSI devices, Paul Gortmaker, (Sun Jan 2, 12:19 am)
Re: [34-longterm 000/260] v2.6.34.8 longterm review, Paul Gortmaker, (Sun Jan 2, 3:46 am)
Re: [34-longterm 000/260] v2.6.34.8 longterm review, Jiri Slaby, (Mon Jan 3, 3:41 am)
Re: [34-longterm 000/260] v2.6.34.8 longterm review, Paul Gortmaker, (Tue Jan 4, 12:11 pm)