Skip to contentSkip to conversation

Fix arguments passed to SHADOW_PRINTK

PATCH In review Yoshiaki Tamura 3 participants 6 messages Sep 11, 2008

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix arguments passed to SHADOW_PRINTK

Hi.

When I compiled xen-3.3-testing with DEBUG_TRACE_DUMP in xen/include/xen/lib.h turned on, I got some errors at SHADOW_PRINTK.
The following patch will fix the arguments passed to SHADOW_PRINTK in xen/arch/x86/mm/shadow/common.c and xen/arch/x86/mm/shadow/multi.c.

Although I haven't tested, it is the same for xen-unstable.hg

Thanks,

Yoshi

xen/arch/x86/mm/shadow/common.c +4 -2
diff -r bfd1157dd315 xen/arch/x86/mm/shadow/common.c
@@ -701,7 +701,8 @@ static void _sh_resync(struct vcpu *v, m
     ASSERT(!sh_page_has_multiple_shadows(mfn_to_page(gmfn)));
 
     SHADOW_PRINTK("d=%d, v=%d, gmfn=%05lx, va=%lx\n",
-                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn), va);
+                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn),
+                  sh_map_domain_page(gmfn));
 
     /* Need to pull write access so the page *stays* in sync. */
     if ( oos_remove_write_access(v, gmfn, fixup) )
@@ -953,7 +954,8 @@ int sh_unsync(struct vcpu *v, mfn_t gmfn
     ASSERT(shadow_locked_by_me(v->domain));
 
     SHADOW_PRINTK("d=%d, v=%d, gmfn=%05lx va %lx\n",
-                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn), va);
+                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn),
+                  sh_map_domain_page(gmfn));
 
     pg = mfn_to_page(gmfn);
 
xen/arch/x86/mm/shadow/multi.c +5 -0
diff -r bfd1157dd315 xen/arch/x86/mm/shadow/multi.c
@@ -3041,9 +3041,14 @@ static int sh_page_fault(struct vcpu *v,
     int fast_emul = 0;
 #endif
 
+#ifdef __x86_64__
     SHADOW_PRINTK("d:v=%u:%u va=%#lx err=%u, rip=%lx\n",
                   v->domain->domain_id, v->vcpu_id, va, regs->error_code,
                   regs->rip);
+#elif defined(__i386__)
+    SHADOW_PRINTK("d:v=%u:%u va=%#lx err=%u\n",
+                  v->domain->domain_id, v->vcpu_id, va, regs->error_code);
+#endif /* __i386__ */
 
     perfc_incr(shadow_fault);
 
Signature · 10 lines
TAMURA, Yoshiaki

NTT Cyber Space Labs
OSS Computing Project
Kernel Group
E-mail: [email protected]
TEL: +81-46-859-2771
FAX: +81-46-855-1152
Address: 1-1 Hikarinooka, Yokosuka
	 Kanagawa 239-0847 JAPAN
#1 lore raw

Hi,

Quoted 6 lines · Yoshiaki Tamura

Hi.

When I compiled xen-3.3-testing with DEBUG_TRACE_DUMP in xen/include/xen/lib.h turned on, I got some errors at SHADOW_PRINTK.
The following patch will fix the arguments passed to SHADOW_PRINTK in xen/arch/x86/mm/shadow/common.c and xen/arch/x86/mm/shadow/multi.c.

My fault, the va argument was removed with recent changes in the OOS code and it is unneeded now.

Quoted 8 lines
     SHADOW_PRINTK("d=%d, v=%d, gmfn=%05lx, va=%lx\n",
xen/arch/x86/mm/shadow/common.c +2 -1
-                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn), va);
+                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn),
+                  sh_map_domain_page(gmfn));
 
     /* Need to pull write access so the page *stays* in sync. */
     if ( oos_remove_write_access(v, gmfn, fixup) )

xen/arch/x86/mm/shadow/common.c int sh_unsync(struct vcpu *v, mfn_t gmfn · line 958

Quoted 6 lines
xen/arch/x86/mm/shadow/common.c +2 -1
@@ -953,7 +954,8 @@ int sh_unsync(struct vcpu *v, mfn_t gmfn
     ASSERT(shadow_locked_by_me(v->domain));
 
     SHADOW_PRINTK("d=%d, v=%d, gmfn=%05lx va %lx\n",
-                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn), va);
+                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn),
+                  sh_map_domain_page(gmfn));

These will likely break Xen in 32 bit, since you should unmap the gmfn in those architecture. As I said, the va argument is unneeded now, so you should remove it.

Thanks,
Gianluca

#2 lore raw
In reply to Yoshiaki Tamura
Quoted 6 lines · Yoshiaki Tamura

When I compiled xen-3.3-testing with DEBUG_TRACE_DUMP in xen/include/xen/lib.h turned on, I got some errors at SHADOW_PRINTK.
The following patch will fix the arguments passed to SHADOW_PRINTK in xen/arch/x86/mm/shadow/common.c and xen/arch/x86/mm/shadow/multi.c.

Although I haven't tested, it is the same for xen-unstable.hg

First chunks of patch: If there is no longer a va to print, then don't manufacture one to print; just remove it entirely from the printk. Apart from anything else, using sh_map_domain_page() without sh_unmap_domain_page() will leak mappings. And the va printed will not be useful anyway.

Final chunk of patch: print regs->eip? That should work on both x86_64 and i386.

Please fix and re-send.

Thanks,
Keir

#3 lore raw

Keir and Gianluca,

Quoted 7 lines · Keir Fraser
Quoted 5 lines · Yoshiaki Tamura

When I compiled xen-3.3-testing with DEBUG_TRACE_DUMP in xen/include/xen/lib.h turned on, I got some errors at SHADOW_PRINTK.
The following patch will fix the arguments passed to SHADOW_PRINTK in xen/arch/x86/mm/shadow/common.c and xen/arch/x86/mm/shadow/multi.c.

Quoted 7 lines
Quoted 1 line

Although I haven't tested, it is the same for xen-unstable.hg

First chunks of patch: If there is no longer a va to print, then don't manufacture one to print; just remove it entirely from the printk. Apart from anything else, using sh_map_domain_page() without sh_unmap_domain_page() will leak mappings. And the va printed will not be useful anyway.

Sorry, that was my mistake.

Quoted 2 lines

Final chunk of patch: print regs->eip? That should work on both x86_64 and i386.

When I compiled xen-3.3-testing with DEBUG_TRACE_DUMP turned on, I got the following errors. Would you please tell me how to work around?

multi.c: In function 'sh_page_fault__guest_2':
multi.c:3045: error: 'struct cpu_user_regs' has no member named 'rip' multi.c: In function 'sh_page_fault__guest_3':
multi.c:3045: error: 'struct cpu_user_regs' has no member named 'rip' make[6]: *** [guest_3.o] Error 1
make[6]: *** Waiting for unfinished jobs....
make[6]: *** [guest_2.o] Error 1
make[5]: *** [shadow/built_in.o] Error 2
make[4]: *** [mm/built_in.o] Error 2
make[3]: *** [/home/tamura/Developer/kemari-v1-devel.hg/xen/arch/x86/built_in.o] Error 2 make[3]: *** Waiting for unfinished jobs....
make[2]: *** [/home/tamura/Developer/kemari-v1-devel.hg/xen/xen] Error 2 make[1]: *** [install] Error 2
make: *** [install-xen] Error 2

Quoted 15 lines

Please fix and re-send.

Thanks,
Keir

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel

Signature · 10 lines
TAMURA, Yoshiaki

NTT Cyber Space Labs
OSS Computing Project
Kernel Group
E-mail: [email protected]
TEL: +81-46-859-2771
FAX: +81-46-855-1152
Address: 1-1 Hikarinooka, Yokosuka
	 Kanagawa 239-0847 JAPAN
#4 lore raw
Quoted 7 lines · Yoshiaki Tamura

Keir and Gianluca,

When I compiled xen-3.3-testing with DEBUG_TRACE_DUMP turned on, I got the following errors. Would you please tell me how to work around?

multi.c: In function 'sh_page_fault__guest_2':
multi.c:3045: error: 'struct cpu_user_regs' has no member named 'rip'

Using regs->eip should work both on 32 and 64 bit.

Thanks,
Gianluca

#5 lore raw

Gianluca,

Thanks for your comment.
The following is the fixed patch.

Thanks,

Yoshi

xen/arch/x86/mm/shadow/common.c +4 -4
diff -r bfd1157dd315 xen/arch/x86/mm/shadow/common.c
@@ -700,8 +700,8 @@ static void _sh_resync(struct vcpu *v, m
              & ~SHF_L1_ANY));
     ASSERT(!sh_page_has_multiple_shadows(mfn_to_page(gmfn)));
 
-    SHADOW_PRINTK("d=%d, v=%d, gmfn=%05lx, va=%lx\n",
-                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn), va);
+    SHADOW_PRINTK("d=%d, v=%d, gmfn=%05lx\n",
+                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn));
 
     /* Need to pull write access so the page *stays* in sync. */
     if ( oos_remove_write_access(v, gmfn, fixup) )
@@ -952,8 +952,8 @@ int sh_unsync(struct vcpu *v, mfn_t gmfn
 
     ASSERT(shadow_locked_by_me(v->domain));
 
-    SHADOW_PRINTK("d=%d, v=%d, gmfn=%05lx va %lx\n",
-                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn), va);
+    SHADOW_PRINTK("d=%d, v=%d, gmfn=%05lx\n",
+                  v->domain->domain_id, v->vcpu_id, mfn_x(gmfn));
 
     pg = mfn_to_page(gmfn);
 
xen/arch/x86/mm/shadow/multi.c +2 -2
diff -r bfd1157dd315 xen/arch/x86/mm/shadow/multi.c
@@ -3041,9 +3041,9 @@ static int sh_page_fault(struct vcpu *v,
     int fast_emul = 0;
 #endif
 
-    SHADOW_PRINTK("d:v=%u:%u va=%#lx err=%u, rip=%lx\n",
+    SHADOW_PRINTK("d:v=%u:%u va=%#lx err=%u, eip=%lx\n",
                   v->domain->domain_id, v->vcpu_id, va, regs->error_code,
-                  regs->rip);
+                  regs->eip);
 
     perfc_incr(shadow_fault);
 
Quoted 24 lines · Gianluca Guida
Quoted 7 lines · Yoshiaki Tamura

Keir and Gianluca,

When I compiled xen-3.3-testing with DEBUG_TRACE_DUMP turned on, I got the following errors. Would you please tell me how to work around?

multi.c: In function 'sh_page_fault__guest_2':
multi.c:3045: error: 'struct cpu_user_regs' has no member named 'rip'

Using regs->eip should work both on 32 and 64 bit.

Thanks,
Gianluca

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel

Signature · 10 lines
TAMURA, Yoshiaki

NTT Cyber Space Labs
OSS Computing Project
Kernel Group
E-mail: [email protected]
TEL: +81-46-859-2771
FAX: +81-46-855-1152
Address: 1-1 Hikarinooka, Yokosuka
	 Kanagawa 239-0847 JAPAN