Re: Patch for debug setcontext

Kumar Gala wrote:

 > I'm guessing, this should really be #if defined(CONFIG_4xx) ||
 > defined(CONFIG_BOOKE)
 > There are a number of similar cases later in this code as well.

That may be the case, but it is very confusing in entry.S.  The dbcr0
register seems to only be defined on 4xx processor, not on book E.
Look, for instance, at load_4xx_dbcr0:, it is only defined if 4xx is
enabled, not if book E is enabled.

-Corey


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

I misspoke a little.  Consider the following code in entry.S:

 #ifdef CONFIG_4xx
    /* Check whether this process has its own DBCR0 value */
    lwz    r0,PTRACE(r2)
    andi.    r0,r0,PT_PTRACED
    bnel-    load_4xx_dbcr0
 #endif

Do we need to add the book E define to this, too?

-Corey

Corey Minyard wrote:

Kumar Gala wrote:

quoted

I'm guessing, this should really be #if defined(CONFIG_4xx) ||
defined(CONFIG_BOOKE)
There are a number of similar cases later in this code as well.

That may be the case, but it is very confusing in entry.S.  The dbcr0
register seems to only be defined on 4xx processor, not on book E.
Look, for instance, at load_4xx_dbcr0:, it is only defined if 4xx is
enabled, not if book E is enabled.

-Corey

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

On Jul 29, 2004, at 9:54 AM, Corey Minyard wrote:

Kumar Gala wrote:

quoted

I'm guessing, this should really be #if defined(CONFIG_4xx) ||
defined(CONFIG_BOOKE)
There are a number of similar cases later in this code as well.

That may be the case, but it is very confusing in entry.S.  The dbcr0
register seems to only be defined on 4xx processor, not on book E.
Look, for instance, at load_4xx_dbcr0:, it is only defined if 4xx is
enabled, not if book E is enabled.

Corey,

I think you may have an old linux-2.6 tree.  I remember renaming
load_4xx_dbcr0 to just load_dbcr0 because of such issues.  Hopefully a
newer tree (one with e500 support, if its going head_e500.S) should be
cleaner.  If not, let me know.

- kumar


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

I'm using stock 2.6.7.  Is there a different kernel I should be using?

-Corey

Kumar Gala wrote:

On Jul 29, 2004, at 9:54 AM, Corey Minyard wrote:

quoted

Kumar Gala wrote:

quoted

I'm guessing, this should really be #if defined(CONFIG_4xx) ||
defined(CONFIG_BOOKE)
There are a number of similar cases later in this code as well.

That may be the case, but it is very confusing in entry.S.  The dbcr0
register seems to only be defined on 4xx processor, not on book E.
Look, for instance, at load_4xx_dbcr0:, it is only defined if 4xx is
enabled, not if book E is enabled.

Corey,

I think you may have an old linux-2.6 tree.  I remember renaming
load_4xx_dbcr0 to just load_dbcr0 because of such issues.  Hopefully a
newer tree (one with e500 support, if its going head_e500.S) should be
cleaner.  If not, let me know.

- kumar

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

If you can look at on of the 2.6.8-rc releases that would be good.  In
adding support for e500, I cleaned up some CONFIG_BOOKE references, and
changed the load_4xx_dbcr0 code to just load_dbcr0 since it applies to
both 4xx and Book-E.

thanks

- kumar

On Jul 29, 2004, at 9:27 PM, Corey Minyard wrote:

I'm using stock 2.6.7.  Is there a different kernel I should be using?

-Corey

Kumar Gala wrote:

quoted

On Jul 29, 2004, at 9:54 AM, Corey Minyard wrote:

quoted

Kumar Gala wrote:

quoted

I'm guessing, this should really be #if defined(CONFIG_4xx) ||
defined(CONFIG_BOOKE)
There are a number of similar cases later in this code as well.

That may be the case, but it is very confusing in entry.S.  The
dbcr0 register seems to only be defined on 4xx processor, not on
book E.  Look, for instance, at load_4xx_dbcr0:, it is only defined
if 4xx is enabled, not if book E is enabled.

Corey,

I think you may have an old linux-2.6 tree.  I remember renaming
load_4xx_dbcr0 to just load_dbcr0 because of such issues.  Hopefully
a newer tree (one with e500 support, if its going head_e500.S) should
be cleaner.  If not, let me know.

- kumar

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Thanks to Kumar, I moved to a new kernel and it makes more sense now.
Patch is attached.

Signed-off-by: Corey Minyard <redacted>