Xen devel - Feb 2007 - [PATCH]mini-os: Fix to get netfront running on ia64

Hi,

to get the netfront stuff on ia64 architecture running I still need to include 
the attached patch.
Please have a look and send comments!
Thanks.

Dietmar.


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

That looks good. Keir could you apply please (if you haven''t done so 
already).

Thanks
Gregor
> Hi,
> 
> to get the netfront stuff on ia64 architecture running I still need to
include
> the attached patch.
> Please have a look and send comments!
> Thanks.
> 
> Dietmar.
> 
> 
> ------------------------------------------------------------------------
> 
> # HG changeset patch
> # User dietmar.hahn@fujitsu-siemens.com
> # Date 1170687542 -3600
> # Node ID 80cd2c0ee40cddec1914542f221159b5911368b3
> # Parent  01ec7dba9ff805a5c74a0318997b747d3e3e3327
> 2 Fixes for supporting netfront on ia64.
> 
> Signed-off-by: Dietmar Hahn <dietmar.hahn@fujitsu-siemens.com>
> 
> 
> diff -r 01ec7dba9ff8 -r 80cd2c0ee40c extras/mini-os/gnttab.c
> --- a/extras/mini-os/gnttab.c	Fri Feb 02 16:07:13 2007 +0000
> +++ b/extras/mini-os/gnttab.c	Mon Feb 05 15:59:02 2007 +0100
> @@ -21,7 +21,12 @@
>  
>  #define NR_RESERVED_ENTRIES 8
>  
> +/* NR_GRANT_FRAMES must be less than or equal to that configured in Xen */
> +#ifdef __ia64__
> +#define NR_GRANT_FRAMES 1
> +#else
>  #define NR_GRANT_FRAMES 4
> +#endif
>  #define NR_GRANT_ENTRIES (NR_GRANT_FRAMES * PAGE_SIZE /
sizeof(grant_entry_t))
>  
>  static grant_entry_t *gnttab_table;
> diff -r 01ec7dba9ff8 -r 80cd2c0ee40c extras/mini-os/netfront.c
> --- a/extras/mini-os/netfront.c	Fri Feb 02 16:07:13 2007 +0000
> +++ b/extras/mini-os/netfront.c	Mon Feb 05 15:59:02 2007 +0100
> @@ -349,7 +349,9 @@ done:
>      init_rx_buffers();
>  
>      unsigned char rawmac[6];
> -    sscanf(mac,"%x:%x:%x:%x:%x:%x",
> +        /* Special conversion specifier ''hh'' needed for
__ia64__. Without
> +           this mini-os panics with ''Unaligned
reference''. */
> +    sscanf(mac,"%hhx:%hhx:%hhx:%hhx:%hhx:%hhx",
>              &rawmac[0],
>              &rawmac[1],
>              &rawmac[2],
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Hi,

this patch added ARCH_LDFLAGS for architecture specific LDFLAGS to the 
makerules and fixed not correct handled rebuild dependencies after changing 
makerule files.
Please have a look and send comments!
Thanks.

Dietmar.


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

That''s a good patch, but I''ve also added a fix to
ARCH_CPFLAGS, and
renamed couple of variables (I didn''t like SPEC_DEPENDS for example).  
Keir could you apply the revised patch?
I hope you don''t mind including you in the Signed-off-by line Dietmar.

Patch info:

- Added ARCH_LDFLAGS for architecture specific LDFLAGS
- Fixed build dependencies after changing makerule files
- Fixed ARCH_CFLAGS for 64bit guest, added ARCH_ASFLAGS
- Couple of variable renames

Signed-off-by: Dietmar Hahn <dietmar.hahn@fujitsu-siemens.com>
Signed-off-by: Grzegorz Milos <gm281@cam.ac.uk>

Thanks
Gregor




On 12 Feb 2007, at 10:36, Dietmar Hahn wrote:
> Hi,
>
> this patch added ARCH_LDFLAGS for architecture specific LDFLAGS to the
> makerules and fixed not correct handled rebuild dependencies after  
> changing
> makerule files.
> Please have a look and send comments!
> Thanks.
>
> Dietmar.
> <mini-os_make.patch>

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

Hi Gregor,

in allocate_xenbus_id() the static variable probe never gets reset.
Therewidth id''s >= NR_REQS are possible, which lead to an overflow
in
req_info[] and may crash the  mini-os.
Thanks.

Dietmar

Signed-off-by: Dietmar Hahn <dietmar.hahn@fujitsu-siemens.com>

# HG changeset patch
# User dietmar.hahn@fujitsu-siemens.com
# Date 1171877953 -3600
# Node ID 3d04558ad3d7e3811ac8c827bb876858bbb1c415
# Parent  b5fc88aad1b0eb35d12e503982c70fdc27f0544a
Because probe never gets decremented (or reset), id >= NR_REQS is possible, 
which may lead to a crash.

diff -r b5fc88aad1b0 -r 3d04558ad3d7 extras/mini-os/xenbus/xenbus.c
--- a/extras/mini-os/xenbus/xenbus.c    Sun Feb 18 15:29:40 2007 +0000
+++ b/extras/mini-os/xenbus/xenbus.c    Mon Feb 19 10:39:13 2007 +0100
@@ -210,7 +210,7 @@ static int allocate_xenbus_id(void)
     }
     nr_live_reqs++;
     req_info[o_probe].in_use = 1;
-    probe = o_probe + 1;
+    probe = (o_probe + 1) % NR_REQS;
     spin_unlock(&req_lock);
     init_waitqueue_head(&req_info[o_probe].waitq);


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

That''s a good catch. Did you see the bug manifesting itself in
practice?

Keir could you apply please?
Thanks.
Gregor

Dietmar Hahn wrote:
> Hi Gregor,
> 
> in allocate_xenbus_id() the static variable probe never gets reset.
> Therewidth id''s >= NR_REQS are possible, which lead to an
overflow in
> req_info[] and may crash the  mini-os.
> Thanks.
> 
> Dietmar
> 
> Signed-off-by: Dietmar Hahn <dietmar.hahn@fujitsu-siemens.com>
> 
> # HG changeset patch
> # User dietmar.hahn@fujitsu-siemens.com
> # Date 1171877953 -3600
> # Node ID 3d04558ad3d7e3811ac8c827bb876858bbb1c415
> # Parent  b5fc88aad1b0eb35d12e503982c70fdc27f0544a
> Because probe never gets decremented (or reset), id >= NR_REQS is
possible,
> which may lead to a crash.
> 
> diff -r b5fc88aad1b0 -r 3d04558ad3d7 extras/mini-os/xenbus/xenbus.c
> --- a/extras/mini-os/xenbus/xenbus.c    Sun Feb 18 15:29:40 2007 +0000
> +++ b/extras/mini-os/xenbus/xenbus.c    Mon Feb 19 10:39:13 2007 +0100
> @@ -210,7 +210,7 @@ static int allocate_xenbus_id(void)
>      }
>      nr_live_reqs++;
>      req_info[o_probe].in_use = 1;
> -    probe = o_probe + 1;
> +    probe = (o_probe + 1) % NR_REQS;
>      spin_unlock(&req_lock);
>      init_waitqueue_head(&req_info[o_probe].waitq);
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Hi Gregor,

Am Dienstag, 20. Februar 2007 19:53 schrieb Grzegorz
Milos:
> That''s a good catch. Did you see the bug manifesting itself in
practice?
>
Yes I got a crash with the mini-os on ia64.
In xenbus_msg_reply() the function allocate_xenbus_id() returns value 33 and 
this leads to a crash in add_waiter().
On x86 this is running fine ;-).

Dietmar.

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