I also used Hirokazu Takahashi's csum patch along with the copy
patch (called as patch2 below) which would use the new copy routine
at the send side also. For the first case where I didn't
use csum patch, the send side used csum_partial_copy_generic.
Since kernel 2.5.38 does not compile for a UNI processor, I ran the test
on SMP with maxcpus set to 1.
Two 996 MHz Pentium III 2-way systems with 4 gig memory/256KB L2 cache
are used. Each system has oneIntel Gigabit Ethernet network card.
2.5.38 kernel from www.kernel.org was used without highmem support.
Intel e1000 driver (the one in the kernel) is used
Napi is not enabled, however I ran the tests with and without TSO.
The TCP_STREAM test was run with TCP-NO Deay ON, 64k socket buffer
size and with a mtu size of 1500. The rest of tcp parameters used are
default values.
Only copypatch is used:
-----------------------
2.5.38 2.5.38+patch -- % improvement using patch --
msg Throughput Throughput Sender Receiver
size 10^6bits/s 10^6bits/s Throughput CPU CPU
512 507.7 538.2 6.0 % -6.4 % 0.0 %
1024 566.3 584.8 3.3 % -16.5 % 1.7 %
2048 503.7 595.8 18.3 % -14.9 % -21.3 %
4096 578.5 600.3 3.8 % -19.7 % -0.5 %
8192 624.7 700.9 12.2 % - 2.2 % -1.4 %
16384 593.5 632.9 6.6 % 0.9 % -6.7 %
32768 659.9 584.9 -11.4 % 7.0 % 11.4 %
65536 655.8 697.7 6.4 % -1.0 % 0.5 %
I used Hirokazu Takahashi's csumpartial_fix patch which is available at
ftp://ftp.valinux.co.jp/pub/people/taka/2.5.36/va-csumpartial-fix-2.5.36.patch
.
The patch applies on 2.5.38 cleanly. With this patch, checksum is done
by the hardware and the new copyroutine is used at the sender side.
copypatch + csumpatch are used:
-------------------------------
2.5.38 2.5.38+patch2 -- % improvement using patch --
msg Throughput Throughput Sender Receiver
size 10^6bits/s 10^6bits/s Throughput CPU CPU
512 507.7 543.0 7.0 % -5.5 % 0.0 %
1024 566.3 587.0 3.7 % -17.8 % 1.5 %
2048 503.7 577.2 14.6 % -19.6 % -11.0 %
4096 578.5 561.0 -3.0 % -2.0 % 4.8 %
8192 624.7 624.2 -0.1 % 11.4 % 10.3 %
16384 593.5 611.0 2.9% 20.5 % -0.4 %
32768 659.9 628.3 -4.8% 27.6 % 7.8 %
65536 655.8 699.6 6.7 % 19.0 % 2.5 %
Since I saw an interrupt pause problem with 2.5.38 running netperf3,
I disabled TSO and ran the same tests with the same configuration
except for TSO and the results from this tests are as follows:
NO_TSO NO_TSO
2.5.38 2.5.38+patch -- % improvement using patch --
msg Throughput Throughput Sender Receiver
size 10^6bits/s 10^6bits/s Throughput CPU CPU
512 497.9 546.3 9.7 % -6.3 % 0.0 %
1024 550.5 603.4 9.6 % -7.5 % 0.0 %
2048 586.9 649.4 10.6 % -8.0% 0.0 %
4096 620.5 690.4 11.3 % -11.6% 0.0 %
8192 647.6 726.8 12.2 % -16.5% 0.0 %
16384 669.1 756.3 13.0 % -14.8% 0.0 %
32768 650.9 734.3 12.8 % -13.4% 0.0 %
65536 626.9 701.8 11.9 % -12.9% 0.1 %
copypatch + csumpatch are used:
-------------------------------
NO_TSO NO_TSO
2.5.38 2.5.38+patch2 -- % improvement using patch --
msg Throughput Throughput Sender Receiver
size 10^6bits/s 10^6bits/s Throughput CPU CPU
512 497.9 540.1 8.5% -5.5 % 0.0 %
1024 550.5 601.2 9.2 % -4.9 % 0.0 %
2048 586.9 647.6 10.3 % -4.7 % 0.0 %
4096 620.5 690.9 11.3 % -0.8 % 0.0 %
8192 647.6 738.3 14.0 % -7.4 % 0.0 %
16384 669.1 763.5 14.1 % -5.3 % 0.0 %
32768 650.9 733.7 12.7 % -3.1 % 0.0 %
65536 626.9 712.8 13.7 % -2.9 % 0.5 %
In the non tso case, the throughput improved for all the message sizes,
and sender side CPU utilization is improved with the csum patch.
I did not see the interrupt pause problem when I disabled TSO in
2.5.38 kernel.
The updated copy patch for 2.5.38 is given here:
diff -Naurb linux-2538/arch/i386/lib/usercopy.c
linux-2538copy/arch/i386/lib/usercopy.c
--- linux-2538/arch/i386/lib/usercopy.c Tue Sep 24 08:25:56 2002
+++ linux-2538copy/arch/i386/lib/usercopy.c Mon Sep 30 10:53:39 2002
@@ -45,8 +45,12 @@
__generic_copy_to_user(void *to, const void *from, unsigned long n)
{
prefetch(from);
- if (access_ok(VERIFY_WRITE, to, n))
+ if (access_ok(VERIFY_WRITE, to, n)){
+ if ((n < 64) || ((((int)from | (int)to) & 0x00000007)==0))
__copy_user(to,from,n);
+ else
+ n=__copy_user_int(to,from,n);
+ }
return n;
}
@@ -54,9 +58,13 @@
__generic_copy_from_user(void *to, const void *from, unsigned long n)
{
prefetchw(to);
- if (access_ok(VERIFY_READ, from, n))
+ if (access_ok(VERIFY_READ, from, n)){
+ if ((n < 64) || ((((int)from | (int)to) & 0x00000007)==0))
__copy_user_zeroing(to,from,n);
else
+ n=__copy_user_zeroing_int(to,from,n);
+ }
+ else
memset(to, 0, n);
return n;
}
@@ -188,3 +196,191 @@
:"cc");
return res & mask;
}
+
+
+/*
+ * Copy To/From Userspace
+ */
+
+/* Generic arbitrary sized copy. */
+unsigned long __copy_user_int(void *to, const void *from,unsigned long
size){
+ int d0, d1;
+ __asm__ __volatile__(
+ " .align 2,0x90\n"
+ "0: movl 32(%4), %%eax\n"
+ " cmpl $67, %0\n"
+ " jbe 1f\n"
+ " movl 64(%4), %%eax\n"
+ " .align 2,0x90\n"
+ "1: movl 0(%4), %%eax\n"
+ " movl 4(%4), %%edx\n"
+ "2: movl %%eax, 0(%3)\n"
+ "21: movl %%edx, 4(%3)\n"
+ " movl 8(%4), %%eax\n"
+ " movl 12(%4),%%edx\n"
+ "3: movl %%eax, 8(%3)\n"
+ "31: movl %%edx, 12(%3)\n"
+ " movl 16(%4), %%eax\n"
+ " movl 20(%4), %%edx\n"
+ "4: movl %%eax, 16(%3)\n"
+ "41: movl %%edx, 20(%3)\n"
+ " movl 24(%4), %%eax\n"
+ " movl 28(%4), %%edx\n"
+ "10: movl %%eax, 24(%3)\n"
+ "51: movl %%edx, 28(%3)\n"
+ " movl 32(%4), %%eax\n"
+ " movl 36(%4), %%edx\n"
+ "11: movl %%eax, 32(%3)\n"
+ "61: movl %%edx, 36(%3)\n"
+ " movl 40(%4), %%eax\n"
+ " movl 44(%4), %%edx\n"
+ "12: movl %%eax, 40(%3)\n"
+ "71: movl %%edx, 44(%3)\n"
+ " movl 48(%4), %%eax\n"
+ " movl 52(%4), %%edx\n"
+ "13: movl %%eax, 48(%3)\n"
+ "81: movl %%edx, 52(%3)\n"
+ " movl 56(%4), %%eax\n"
+ " movl 60(%4), %%edx\n"
+ "14: movl %%eax, 56(%3)\n"
+ "91: movl %%edx, 60(%3)\n"
+ " addl $-64, %0\n"
+ " addl $64, %4\n"
+ " addl $64, %3\n"
+ " cmpl $63, %0\n"
+ " ja 0b\n"
+ "5: movl %0, %%eax\n"
+ " shrl $2, %0\n"
+ " andl $3, %%eax\n"
+ " cld\n"
+ "6: rep; movsl\n"
+ " movl %%eax, %0\n"
+ "7: rep; movsb\n"
+ "8:\n"
+ ".section .fixup,\"ax\"\n"
+ "9: lea 0(%%eax,%0,4),%0\n"
+ " jmp 8b\n"
+ "15: movl %6, %0\n"
+ " jmp 8b\n"
+ ".previous\n"
+ ".section __ex_table,\"a\"\n"
+ " .align 4\n"
+ " .long 2b,15b\n"
+ " .long 21b,15b\n"
+ " .long 3b,15b\n"
+ " .long 31b,15b\n"
+ " .long 4b,15b\n"
+ " .long 41b,15b\n"
+ " .long 10b,15b\n"
+ " .long 51b,15b\n"
+ " .long 11b,15b\n"
+ " .long 61b,15b\n"
+ " .long 12b,15b\n"
+ " .long 71b,15b\n"
+ " .long 13b,15b\n"
+ " .long 81b,15b\n"
+ " .long 14b,15b\n"
+ " .long 91b,15b\n"
+ " .long 6b,9b\n"
+ " .long 7b,8b\n"
+ ".previous"
+ : "=&c"(size), "=&D" (d0), "=&S" (d1)
+ : "1"(to), "2"(from), "0"(size),"i"(-EFAULT)
+ : "eax", "edx", "memory");
+ return size;
+}
+
+unsigned long __copy_user_zeroing_int(void *to,const void *from, unsigned
long size){
+ int d0, d1;
+ __asm__ __volatile__(
+ " .align 2,0x90\n"
+ "0: movl 32(%4), %%eax\n"
+ " cmpl $67, %0\n"
+ " jbe 2f\n"
+ "1: movl 64(%4), %%eax\n"
+ " .align 2,0x90\n"
+ "2: movl 0(%4), %%eax\n"
+ "21: movl 4(%4), %%edx\n"
+ " movl %%eax, 0(%3)\n"
+ " movl %%edx, 4(%3)\n"
+ "3: movl 8(%4), %%eax\n"
+ "31: movl 12(%4),%%edx\n"
+ " movl %%eax, 8(%3)\n"
+ " movl %%edx, 12(%3)\n"
+ "4: movl 16(%4), %%eax\n"
+ "41: movl 20(%4), %%edx\n"
+ " movl %%eax, 16(%3)\n"
+ " movl %%edx, 20(%3)\n"
+ "10: movl 24(%4), %%eax\n"
+ "51: movl 28(%4), %%edx\n"
+ " movl %%eax, 24(%3)\n"
+ " movl %%edx, 28(%3)\n"
+ "11: movl 32(%4), %%eax\n"
+ "61: movl 36(%4), %%edx\n"
+ " movl %%eax, 32(%3)\n"
+ " movl %%edx, 36(%3)\n"
+ "12: movl 40(%4), %%eax\n"
+ "71: movl 44(%4), %%edx\n"
+ " movl %%eax, 40(%3)\n"
+ " movl %%edx, 44(%3)\n"
+ "13: movl 48(%4), %%eax\n"
+ "81: movl 52(%4), %%edx\n"
+ " movl %%eax, 48(%3)\n"
+ " movl %%edx, 52(%3)\n"
+ "14: movl 56(%4), %%eax\n"
+ "91: movl 60(%4), %%edx\n"
+ " movl %%eax, 56(%3)\n"
+ " movl %%edx, 60(%3)\n"
+ " addl $-64, %0\n"
+ " addl $64, %4\n"
+ " addl $64, %3\n"
+ " cmpl $63, %0\n"
+ " ja 0b\n"
+ "5: movl %0, %%eax\n"
+ " shrl $2, %0\n"
+ " andl $3, %%eax\n"
+ " cld\n"
+ "6: rep; movsl\n"
+ " movl %%eax,%0\n"
+ "7: rep; movsb\n"
+ "8:\n"
+ ".section .fixup,\"ax\"\n"
+ "9: lea 0(%%eax,%0,4),%0\n"
+ "16: pushl %0\n"
+ " pushl %%eax\n"
+ " xorl %%eax,%%eax\n"
+ " rep; stosb\n"
+ " popl %%eax\n"
+ " popl %0\n"
+ " jmp 8b\n"
+ "15: movl %6, %0\n"
+ " jmp 8b\n"
+ ".previous\n"
+ ".section __ex_table,\"a\"\n"
+ " .align 4\n"
+ " .long 0b,16b\n"
+ " .long 1b,16b\n"
+ " .long 2b,16b\n"
+ " .long 21b,16b\n"
+ " .long 3b,16b\n"
+ " .long 31b,16b\n"
+ " .long 4b,16b\n"
+ " .long 41b,16b\n"
+ " .long 10b,16b\n"
+ " .long 51b,16b\n"
+ " .long 11b,16b\n"
+ " .long 61b,16b\n"
+ " .long 12b,16b\n"
+ " .long 71b,16b\n"
+ " .long 13b,16b\n"
+ " .long 81b,16b\n"
+ " .long 14b,16b\n"
+ " .long 91b,16b\n"
+ " .long 6b,9b\n"
+ " .long 7b,16b\n"
+ ".previous"
+ : "=&c"(size), "=&D" (d0), "=&S" (d1)
+ : "1"(to), "2"(from), "0"(size),"i"(-EFAULT)
+ : "eax", "edx", "memory");
+ return size;
+}
diff -Naurb linux-2538/include/asm-i386/uaccess.h
linux-2538copy/include/asm-i386/uaccess.h
--- linux-2538/include/asm-i386/uaccess.h Thu Sep 26 16:27:47 2002
+++ linux-2538copy/include/asm-i386/uaccess.h Fri Sep 27 16:10:14 2002
@@ -34,6 +34,8 @@
#define segment_eq(a,b) ((a).seg == (b).seg)
extern int __verify_write(const void *, unsigned long);
+extern unsigned long __copy_user_int(void *to, const void *from, unsigned
long size);
+extern unsigned long __copy_user_zeroing_int(void *to, const void *from,
unsigned long size);
#define __addr_ok(addr) ((unsigned long)(addr) < (current_thread_info()
->addr_limit.seg))
@@ -309,14 +311,20 @@
static inline unsigned long
__generic_copy_from_user_nocheck(void *to, const void *from, unsigned long
n)
{
+ if ((n < 64) || ((((int)from | (int)to) & 0x00000007)==0))
__copy_user_zeroing(to,from,n);
+ else
+ n=__copy_user_zeroing_int(to,from,n);
return n;
}
static inline unsigned long
__generic_copy_to_user_nocheck(void *to, const void *from, unsigned long
n)
{
+ if ((n < 64) || ((((int)from | (int)to) & 0x00000007)==0))
__copy_user(to,from,n);
+ else
+ n=__copy_user_int(to,from,n);
return n;
}
@@ -578,24 +586,16 @@
}
#define copy_to_user(to,from,n) \
- (__builtin_constant_p(n) ? \
- __constant_copy_to_user((to),(from),(n)) : \
- __generic_copy_to_user((to),(from),(n)))
+ __generic_copy_to_user((to),(from),(n))
#define copy_from_user(to,from,n) \
- (__builtin_constant_p(n) ? \
- __constant_copy_from_user((to),(from),(n)) : \
- __generic_copy_from_user((to),(from),(n)))
+ __generic_copy_from_user((to),(from),(n))
#define __copy_to_user(to,from,n) \
- (__builtin_constant_p(n) ? \
- __constant_copy_to_user_nocheck((to),(from),(n)) : \
- __generic_copy_to_user_nocheck((to),(from),(n)))
+ __generic_copy_to_user_nocheck((to),(from),(n))
#define __copy_from_user(to,from,n) \
- (__builtin_constant_p(n) ? \
- __constant_copy_from_user_nocheck((to),(from),(n)) : \
- __generic_copy_from_user_nocheck((to),(from),(n)))
+ __generic_copy_from_user_nocheck((to),(from),(n))
long strncpy_from_user(char *dst, const char *src, long count);
long __strncpy_from_user(char *dst, const char *src, long count);
Regards,
Mala
Mala Anand
IBM Linux Technology Center - Kernel Performance
E-mail:manand@us.ibm.com
http://www-124.ibm.com/developerworks/opensource/linuxperf
http://www-124.ibm.com/developerworks/projects/linuxperf
Phone:838-8088; Tie-line:678-8088
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/