[PATCH] [SMP] __GENERIC_PER_CPU changes

Now CONFIG_DEBUG_INITDATA is in, initial percpu data [__per_cpu_start,__per_cpu_end] can be declared as a redzone, and invalid accesses after boot can be detected, at least for i386.

We can let non possible cpus percpu data point to this 'redzone' instead of NULL .

NULL was not a good choice because part of [0..32768] memory may be readable and invalid accesses may happen unnoticed.

If CONFIG_DEBUG_INITDATA is not defined, each non possible cpu points to the initial percpu data (__per_cpu_offset[cpu] == 0), thus invalid accesses wont be detected/crash.

This patch also moves __per_cpu_offset[] to read_mostly area to avoid false sharing.

Signed-off-by: Eric Dumazet <[email protected]>

--- a/init/main.c	2006-02-01 10:44:10.000000000 +0100
+++ b/init/main.c	2006-02-01 10:50:16.000000000 +0100
@@ -325,7 +325,7 @@
 #else
 
 #ifdef __GENERIC_PER_CPU
-unsigned long __per_cpu_offset[NR_CPUS];
+unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
 
 EXPORT_SYMBOL(__per_cpu_offset);
 
@@ -343,11 +343,7 @@
 #endif
 	ptr = alloc_bootmem(size * nr_possible_cpus);
 
-	for (i = 0; i < NR_CPUS; i++) {
-		if (!cpu_possible(i)) {
-			__per_cpu_offset[i] = (char*)0 - __per_cpu_start;
-			continue;
-		}
+	for_each_cpu(i) {
 		__per_cpu_offset[i] = ptr - __per_cpu_start;
 		memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
 		ptr += size;

---

• References:
  • [PATCH] tvec_bases too large for per-cpu data
    • From: "Jan Beulich" <[email protected]>
  • Re: [PATCH] tvec_bases too large for per-cpu data
    • From: Andrew Morton <[email protected]>
  • Re: [PATCH] tvec_bases too large for per-cpu data
    • From: "Jan Beulich" <[email protected]>
  • Re: [PATCH] tvec_bases too large for per-cpu data
    • From: Andrew Morton <[email protected]>
  • Re: [PATCH] tvec_bases too large for per-cpu data
    • From: "Jan Beulich" <[email protected]>
  • Re: [PATCH] tvec_bases too large for per-cpu data
    • From: Andrew Morton <[email protected]>
  • [PATCH] [SMP] reduce size of percpudata, and make sure per_cpu(object, not_possible_cpu) cause an invalid memory reference
    • From: Eric Dumazet <[email protected]>

• Prev by Date: [patch 04/44] use non atomic operations for minix_*_bit() and ext2_*_bit()
• Next by Date: [patch 06/44] generic __{,test_and_}{set,clear,change}_bit() and test_bit()
• Previous by thread: Re: [PATCH] [SMP] reduce size of percpudata, and make sure per_cpu(object, not_possible_cpu) cause an invalid memory reference
• Next by thread: Re: [PATCH] tvec_bases too large for per-cpu data
• Index(es):
  • Date
  • Thread

[Index of Archives]     [Kernel Newbies]     [Netfilter]     [Bugtraq]     [Photo]     [Stuff]     [Gimp]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Video 4 Linux]     [Linux for the blind]     [Linux Resources]