[Adeos-main] [Resent][RFC-PATCH] Instrument smp_processor_id /wrt illicit usage

[Adeos-main] [Resent][RFC-PATCH] Instrument smp_processor_id /wrt illicit usage

[Jan Kiszka]

[-- Attachment #1.1: Type: text/plain, Size: 452 bytes --]

This is still in my local queue, don't recall anymore when we last
discussed it and if there was some issue remaining.

Anyway, here is a rebased set to detect calls of smp_processor_id() from
illicit contexts (non-root) on archs that bother about this. Only i386
is cared about so far, x86_64 and PowerPC should be treated similarly,
other (future) SMP-arch would arm the check automatically. Yet another
safety bag for porting I-pipe.

Jan

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.2: instrument-smp_processor_id.patch --]
[-- Type: text/x-patch; name="instrument-smp_processor_id.patch", Size: 1238 bytes --]

---
 lib/smp_processor_id.c |   20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)

Index: linux-2.6.22-ipipe/lib/smp_processor_id.c
===================================================================
--- linux-2.6.22-ipipe.orig/lib/smp_processor_id.c
+++ linux-2.6.22-ipipe/lib/smp_processor_id.c
@@ -10,11 +10,27 @@
 unsigned int debug_smp_processor_id(void)
 {
 	unsigned long preempt_count = preempt_count();
-	int this_cpu = raw_smp_processor_id();
+	int this_cpu = ipipe_processor_id();
 	cpumask_t this_mask;
 
-	if (!ipipe_root_domain_p)
+#ifdef CONFIG_IPIPE
+	if (!ipipe_root_domain_p) {
+#ifndef IPIPE_STACK_INVARIANT_CPUID
+		if (test_bit(IPIPE_NOSTACK_FLAG, &ipipe_current_domain->
+						 cpudata[this_cpu].status)) {
+			ipipe_trace_panic_freeze();
+			printk(KERN_ERR "BUG: using smp_processor_id() over "
+			       "non-Linux stack. I-pipe domain: %s\n",
+			       ipipe_current_domain->name);
+			print_symbol("caller is %s\n",
+				     (long)__builtin_return_address(0));
+			dump_stack();
+			ipipe_trace_panic_dump();
+		}
+#endif /* !IPIPE_STACK_INVARIANT_CPUID */
 		goto out;
+	}
+#endif /* CONFIG_IPIPE */
 
 	if (likely(preempt_count))
 		goto out;

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1.3: instrument-smp_processor_id-i386.patch --]
[-- Type: text/x-patch; name="instrument-smp_processor_id-i386.patch", Size: 498 bytes --]

---
 include/asm-i386/ipipe.h |    1 +
 1 file changed, 1 insertion(+)

Index: linux-2.6.22-ipipe/include/asm-i386/ipipe.h
===================================================================
--- linux-2.6.22-ipipe.orig/include/asm-i386/ipipe.h
+++ linux-2.6.22-ipipe/include/asm-i386/ipipe.h
@@ -38,6 +38,7 @@
 #include <asm/ptrace.h>
 
 #define ipipe_processor_id()   raw_smp_processor_id()
+#define IPIPE_STACK_INVARIANT_CPUID
 
 #define prepare_arch_switch(next)		\
 do {						\

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 249 bytes --]

Re: [Adeos-main] [Resent][RFC-PATCH] Instrument smp_processor_id /wrt illicit usage

[Philippe Gerum]

On Sat, 2007-09-15 at 09:51 +0200, Jan Kiszka wrote:
> This is still in my local queue, don't recall anymore when we last
> discussed it and if there was some issue remaining.
> 
> Anyway, here is a rebased set to detect calls of smp_processor_id() from
> illicit contexts (non-root) on archs that bother about this. Only i386
> is cared about so far, x86_64 and PowerPC should be treated similarly,
> other (future) SMP-arch would arm the check automatically. Yet another
> safety bag for porting I-pipe.

Last time, the ancestor of this patch badly broke my x86_84 setup,
causing hell in the NMI path for instance. A number of fixes you sent
lately have improved the situation since then for sure, so I will hold
those patches until the pending 2.6.22 issues are fixed, but will merge
in the end.

> 
> Jan
-- 
Philippe.