2014-06-03 16:23:49

by Petr Mladek

Subject: [PATCH] ftrace/x86: Call text_ip_addr() instead of the duplicated code

I just went over this when looking at some Xen-related ftrace initialization
problems. They were related to Xen code that is not upstream but this clean up
would make sense here.

I think that this was already the intention when text_ip_addr() was introduced
in the commit 87fbb2ac6073a703930 (ftrace/x86: Use breakpoints for converting
function graph caller). Anyway, better do it now before it shots people into
their leg ;-)

Signed-off-by: Petr Mladek <[email protected]>
---
arch/x86/kernel/ftrace.c | 11 +----------
1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c
index 89de3eaf8772..cbc4a91b131e 100644
--- a/arch/x86/kernel/ftrace.c
+++ b/arch/x86/kernel/ftrace.c
@@ -297,16 +297,7 @@ int ftrace_int3_handler(struct pt_regs *regs)

static int ftrace_write(unsigned long ip, const char *val, int size)
{
- /*
- * On x86_64, kernel text mappings are mapped read-only with
- * CONFIG_DEBUG_RODATA. So we use the kernel identity mapping instead
- * of the kernel text mapping to modify the kernel text.
- *
- * For 32bit kernels, these mappings are same and we can use
- * kernel identity mapping to modify code.
- */
- if (within(ip, (unsigned long)_text, (unsigned long)_etext))
- ip = (unsigned long)__va(__pa_symbol(ip));
+ ip = text_ip_addr(ip);

if (probe_kernel_write((void *)ip, val, size))
return -EPERM;
--
1.8.4

2014-06-03 23:01:55

by Steven Rostedt

Subject: Re: [PATCH] ftrace/x86: Call text_ip_addr() instead of the duplicated code

On Tue, 3 Jun 2014 18:23:21 +0200
Petr Mladek <[email protected]> wrote:

> I just went over this when looking at some Xen-related ftrace initialization
> problems. They were related to Xen code that is not upstream but this clean up
> would make sense here.
>
> I think that this was already the intention when text_ip_addr() was introduced
> in the commit 87fbb2ac6073a703930 (ftrace/x86: Use breakpoints for converting
> function graph caller). Anyway, better do it now before it shots people into
> their leg ;-)

You know, I made this change. It somehow got lost :-/

I better analyze this a bit to make sure it was just lost and I didn't
find some odd ball reason not to apply it. But I doubt I'll find
anything. If not, I'll get it enqueued for 3.16.

Thanks!

-- Steve


>
> Signed-off-by: Petr Mladek <[email protected]>
> ---
> arch/x86/kernel/ftrace.c | 11 +----------
> 1 file changed, 1 insertion(+), 10 deletions(-)
>
> diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c
> index 89de3eaf8772..cbc4a91b131e 100644
> --- a/arch/x86/kernel/ftrace.c
> +++ b/arch/x86/kernel/ftrace.c
> @@ -297,16 +297,7 @@ int ftrace_int3_handler(struct pt_regs *regs)
>
> static int ftrace_write(unsigned long ip, const char *val, int size)
> {
> - /*
> - * On x86_64, kernel text mappings are mapped read-only with
> - * CONFIG_DEBUG_RODATA. So we use the kernel identity mapping instead
> - * of the kernel text mapping to modify the kernel text.
> - *
> - * For 32bit kernels, these mappings are same and we can use
> - * kernel identity mapping to modify code.
> - */
> - if (within(ip, (unsigned long)_text, (unsigned long)_etext))
> - ip = (unsigned long)__va(__pa_symbol(ip));
> + ip = text_ip_addr(ip);
>
> if (probe_kernel_write((void *)ip, val, size))
> return -EPERM;