[PATCH] Mini-os get_current() fixup

[PATCH] Mini-os get_current() fixup

[Samuel Thibault]

[-- Attachment #1: Type: text/plain, Size: 527 bytes --]

Hi,

I'm currently playing with mini-os, and noticed bogus code:

0x00019301 <do_page_fault+81>:  mov    $0x1,%ecx
0x00019306 <do_page_fault+86>:  mov    $0xffffe000,%eax
0x0001930b <do_page_fault+91>:  mov    %ecx,0x32980
0x00019311 <do_page_fault+97>:  and    %esp,%edx
0x00019313 <do_page_fault+99>:  mov    (%edx),%ecx
0x00019315 <do_page_fault+101>: mov    (%ecx),%eax

Which is supposed to approximately correspond to get_current(). It
happens that %edx is never set to 0xfffff000. The attached patch fixes
that.

Samuel

[-- Attachment #2: patch --]
[-- Type: text/plain, Size: 623 bytes --]

diff -r 3746b3d4f301 extras/mini-os/include/x86/arch_sched.h
--- a/extras/mini-os/include/x86/arch_sched.h	Thu Feb 22 15:26:21 2007 +0000
+++ b/extras/mini-os/include/x86/arch_sched.h	Mon Oct 15 17:23:01 2007 +0100
@@ -7,9 +7,9 @@ static inline struct thread* get_current
 {
     struct thread **current;
 #ifdef __i386__    
-    __asm__("andl %%esp,%0; ":"=r" (current) : "r" (~8191UL));
+    __asm__("andl %%esp,%0; ":"=r" (current) : "0" (~8191UL));
 #else
-    __asm__("andq %%rsp,%0; ":"=r" (current) : "r" (~8191UL));
+    __asm__("andq %%rsp,%0; ":"=r" (current) : "0" (~8191UL));
 #endif 
     return *current;
 }

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: [PATCH] Mini-os get_current() fixup

[Samuel Thibault]

Samuel Thibault, le Mon 15 Oct 2007 17:23:32 +0100, a écrit :
> 0x00019301 <do_page_fault+81>:  mov    $0x1,%ecx
> 0x00019306 <do_page_fault+86>:  mov    $0xffffe000,%eax
> 0x0001930b <do_page_fault+91>:  mov    %ecx,0x32980
> 0x00019311 <do_page_fault+97>:  and    %esp,%edx
> 0x00019313 <do_page_fault+99>:  mov    (%edx),%ecx
> 0x00019315 <do_page_fault+101>: mov    (%ecx),%eax
> 
> Which is supposed to approximately correspond to get_current(). It
> happens that %edx is never set to 0xfffff000. The attached patch fixes

(I meant 0xffffe000 of course)

Samuel

Re: [PATCH] Mini-os xenbus wait fixup

[Samuel Thibault]

[-- Attachment #1: Type: text/plain, Size: 258 bytes --]

Hi,

I'm getting spurious wakes and crashes when using xenbus_wake_for_value.
It looks to me like the attached patch is needed because else the wait
element is kept in the watch queue (just like in the wait_event() macro,
and it does fix my issues).

Samuel

[-- Attachment #2: patch --]
[-- Type: text/plain, Size: 390 bytes --]

diff -r b28ae5f00553 extras/mini-os/xenbus/xenbus.c
--- a/extras/mini-os/xenbus/xenbus.c	Tue Oct 23 09:26:43 2007 +0100
+++ b/extras/mini-os/xenbus/xenbus.c	Wed Oct 24 14:27:10 2007 +0100
@@ -79,6 +79,7 @@ void wait_for_watch(void)
     schedule();
     remove_waiter(w);
     wake(current);
+    remove_wait_queue(&w);
 }
 
 char* xenbus_wait_for_value(const char* path,const char* value)

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel