Re: Potato woes

[Loic Prylli <Loic.Prylli@ens-lyon.fr>]

Nikita Schmidt writes:
 > [1  <text/plain; us-ascii (7bit)>]
 > On Monday, 26 Apr, Giuliano P Procida wrote:
 > > 
 > > In init (sometimes), login, klogd (really bad this one). Here is an
 > > example oops (I used -c 4 as the -c 1 output looked dubious):
 > > 
 > 
 > That looks slightly familiar.  Try the patch attached (courtesy of Richard
 > Henderson).
 > 

Just for my own education, can anyone explain why loading the gp is
required in the definition of copy_user.S, and strncpy_from_user.S ? 
(I perfectly see why the other changes are necessary, especially in
the assembler macros to force loading the gp when a subroutine called
is made). But in copy_user.S, the comment is "# we do exceptions -- we
need the gp" , but I have not found where in the exception code the gp
register of the possibly trapping code is used.

Besides, does somebody know to whom should I submit my previous two
lines adjtimex patch, as I cannot follow axp-list neither
linux-kernel, I do not feel like sending it there.

Regards,

Loic


 > diff -rup 2.2.6-dist/arch/alpha/lib/copy_user.S 2.2.6/arch/alpha/lib/copy_user.S
 > --- 2.2.6-dist/arch/alpha/lib/copy_user.S	Sat Apr 25 22:35:18 1998
 > +++ 2.2.6/arch/alpha/lib/copy_user.S	Sat Apr 24 15:13:19 1999
 > @@ -46,6 +46,8 @@
 >  	.globl __copy_user
 >  	.ent __copy_user
 >  __copy_user:
 > +	ldgp $29,0($27)			# we do exceptions -- we need the gp.
 > +	.prologue 1
 >  	and $6,7,$3
 >  	beq $0,$35
 >  	beq $3,$36
 > ...