>> void alternatives_smp_module_del(struct module *mod)
>> {
>> struct smp_alt_module *item;
>> -unsigned long flags;
>>
>> if (smp_alt_once || noreplace_smp)
>> return;
>>
>> -spin_lock_irqsave(&smp_alt, flags);
>> +spin_lock(&smp_alt);
>
>Unrelated change?
Hi Andi,
I guess the patch I proposed there:
"Text Edit Lock"
http://www.ussg.iu.edu/hypermail/linux/kernel/0707.1/2980.html
solves the problem more generally than the patch discussed in this
thread. It provides locking (architecture independent) and correct
page protection (architecture dependen
Jan Beulich wrote:
> + if (va < MODULES_VADDR) {
> + change_page_attr(virt_to_page(instr),
> + PFN_UP(va + len) - PFN_DOWN(va),
> +#ifdef CONFIG_X86_64
> + PAGE_KERNEL_RO);
> +#else
> + PAGE_KERNE
On Wednesday 04 July 2007 08:33:24 Jan Beulich wrote:
Sorry for late feedback.
> +#ifdef CONFIG_DEBUG_RODATA
> +
> +#ifdef CONFIG_X86_32
> +#include
> +#define MODULES_VADDR VMALLOC_START
> +#endif
> +
> +static inline void make_writable(const void *instr, unsigned int len)
> +{
> + unsigne
Dave Jones wrote:
On Wed, Jul 04, 2007 at 08:41:39AM -0700, Jeremy Fitzhardinge wrote:
> I'd already posted a patch to remove smp alternatives. I thought it had
> been merged already, but I guess it's in Andi's queue.
What was the rationale behind that? Given many distros have now
moved
On Wed, Jul 04, 2007 at 08:41:39AM -0700, Jeremy Fitzhardinge wrote:
> I'd already posted a patch to remove smp alternatives. I thought it had
> been merged already, but I guess it's in Andi's queue.
What was the rationale behind that? Given many distros have now
moved to shipping just an S
Jan Beulich wrote:
Oh, sorry, I forgot to remove that part of the comment after merging up to
2.6..22-rcX.
Actually, no, the comment is right - it talks about pieces left in x86-64's
vmlinux.lds.S.
Hm. I was pretty sure I did that too, but maybe it was in a separate
patch which got los
Andi Kleen wrote:
Not if the patching happens early enough at boot. That's true for Xen
and lguest, but I'm not sure about VMI
paravirt_ops patching happens at the same time as all other patching,
via check_bugs(). So its the same regardless of backend, and quite late.
J
-
To unsubscrib
>>> On x86-64 additionally remove all mappings past the kernel image, and
>>> remove leftovers from the removal of the more general (but abandoned)
>>> SMP alternatives.
>>>
>>
>>I'd already posted a patch to remove smp alternatives. I thought it had
>>been merged already, but I guess it's in
>>> Jeremy Fitzhardinge <[EMAIL PROTECTED]> 04.07.07 17:41 >>>
>Jan Beulich wrote:
>> Instead of suppressing the change of .text to become readonly, make
>> the SMP locks patching code properly adjust/restore the page access
>> rights.
>>
>
>I guess we'll need to do the same thing for paravirt_o
On Wednesday 04 July 2007 17:41:39 Jeremy Fitzhardinge wrote:
> Jan Beulich wrote:
> > Instead of suppressing the change of .text to become readonly, make
> > the SMP locks patching code properly adjust/restore the page access
> > rights.
> >
>
> I guess we'll need to do the same thing for para
Jan Beulich wrote:
Instead of suppressing the change of .text to become readonly, make
the SMP locks patching code properly adjust/restore the page access
rights.
I guess we'll need to do the same thing for paravirt_ops patching?
On x86-64 additionally remove all mappings past the kernel i
Instead of suppressing the change of .text to become readonly, make
the SMP locks patching code properly adjust/restore the page access
rights.
On x86-64 additionally remove all mappings past the kernel image, and
remove leftovers from the removal of the more general (but abandoned)
SMP alternativ
13 matches
Mail list logo
