Currently, the log-level of show_stack() depends on a platform
realization. It creates situations where the headers are printed with
lower log level or higher than the stacktrace (depending on
a platform or user).
Furthermore, it forces the logic decision from user to an architecture
side. In result, some users as sysrq/kdb/etc are doing tricks with
temporary rising console_loglevel while printing their messages.
And in result it not only may print unwanted messages from other CPUs,
but also omit printing at all in the unlucky case where the printk()
was deferred.
Introducing log-level parameter and KERN_UNSUPPRESSED [1] seems
an easier approach than introducing more printk buffers.
Also, it will consolidate printings with headers.
Add log level argument to printk_address() as a preparation to introduce
show_stack_loglvl().
As a good side-effect show_fault_oops() now prints the address with
KERN_EMREG as the rest of output, making sure there won't be situation
where "PC: " is printed without actual address.
Cc: Rich Felker <[email protected]>
Cc: [email protected]
[1]: https://lore.kernel.org/lkml/[email protected]/T/#u
Signed-off-by: Dmitry Safonov <[email protected]>
---
arch/sh/include/asm/kdebug.h | 3 ++-
arch/sh/kernel/dumpstack.c | 6 +++---
arch/sh/mm/fault.c | 2 +-
3 files changed, 6 insertions(+), 5 deletions(-)
diff --git a/arch/sh/include/asm/kdebug.h b/arch/sh/include/asm/kdebug.h
index de8693fabb1d..960545306afa 100644
--- a/arch/sh/include/asm/kdebug.h
+++ b/arch/sh/include/asm/kdebug.h
@@ -12,7 +12,8 @@ enum die_val {
};
/* arch/sh/kernel/dumpstack.c */
-extern void printk_address(unsigned long address, int reliable);
+extern void printk_address(unsigned long address, int reliable,
+ const char *loglvl);
extern void dump_mem(const char *str, const char *loglvl,
unsigned long bottom, unsigned long top);
diff --git a/arch/sh/kernel/dumpstack.c b/arch/sh/kernel/dumpstack.c
index 2c1a78e5776b..959064b90055 100644
--- a/arch/sh/kernel/dumpstack.c
+++ b/arch/sh/kernel/dumpstack.c
@@ -44,9 +44,9 @@ void dump_mem(const char *str, const char *loglvl,
}
}
-void printk_address(unsigned long address, int reliable)
+void printk_address(unsigned long address, int reliable, const char *loglvl)
{
- printk(" [<%p>] %s%pS\n", (void *) address,
+ printk("%s [<%p>] %s%pS\n", loglvl, (void *) address,
reliable ? "" : "? ", (void *) address);
}
@@ -118,7 +118,7 @@ static int print_trace_stack(void *data, char *name)
*/
static void print_trace_address(void *data, unsigned long addr, int reliable)
{
- printk_address(addr, reliable);
+ printk_address(addr, reliable, (char *)data);
}
static const struct stacktrace_ops print_trace_ops = {
diff --git a/arch/sh/mm/fault.c b/arch/sh/mm/fault.c
index 5f23d7907597..f5da8f5ea389 100644
--- a/arch/sh/mm/fault.c
+++ b/arch/sh/mm/fault.c
@@ -196,7 +196,7 @@ show_fault_oops(struct pt_regs *regs, unsigned long address)
printk(KERN_CONT " at %08lx\n", address);
printk(KERN_ALERT "PC:");
- printk_address(regs->pc, 1);
+ printk_address(regs->pc, 1, KERN_ALERT);
show_pte(NULL, address);
}
--
2.26.0
On Sat, 18 Apr 2020 21:19:25 +0100 Dmitry Safonov <[email protected]> wrote:
> Currently, the log-level of show_stack() depends on a platform
> realization. It creates situations where the headers are printed with
> lower log level or higher than the stacktrace (depending on
> a platform or user).
>
> Furthermore, it forces the logic decision from user to an architecture
> side. In result, some users as sysrq/kdb/etc are doing tricks with
> temporary rising console_loglevel while printing their messages.
> And in result it not only may print unwanted messages from other CPUs,
> but also omit printing at all in the unlucky case where the printk()
> was deferred.
>
> Introducing log-level parameter and KERN_UNSUPPRESSED [1] seems
> an easier approach than introducing more printk buffers.
> Also, it will consolidate printings with headers.
>
> Add log level argument to printk_address() as a preparation to introduce
> show_stack_loglvl().
>
> As a good side-effect show_fault_oops() now prints the address with
> KERN_EMREG as the rest of output, making sure there won't be situation
> where "PC: " is printed without actual address.
>
> --- a/arch/sh/include/asm/kdebug.h
> +++ b/arch/sh/include/asm/kdebug.h
> @@ -12,7 +12,8 @@ enum die_val {
> };
>
> /* arch/sh/kernel/dumpstack.c */
> -extern void printk_address(unsigned long address, int reliable);
> +extern void printk_address(unsigned long address, int reliable,
> + const char *loglvl);
> extern void dump_mem(const char *str, const char *loglvl,
> unsigned long bottom, unsigned long top);
>
> ...
>
> --- a/arch/sh/mm/fault.c
> +++ b/arch/sh/mm/fault.c
> @@ -196,7 +196,7 @@ show_fault_oops(struct pt_regs *regs, unsigned long address)
>
> printk(KERN_CONT " at %08lx\n", address);
> printk(KERN_ALERT "PC:");
> - printk_address(regs->pc, 1);
> + printk_address(regs->pc, 1, KERN_ALERT);
>
It would be more intuitive to do
printk_address(KERN_ALERT, regs->pc, 1);
because the loglevel always comes first.
I guess it doesn't matter much, as sh seems to be rather dead.
On 4/20/20 5:52 PM, Andrew Morton wrote:
> I guess it doesn't matter much, as sh seems to be rather dead.
The j-core guys are using it.
Rob
On Sat, Apr 25, 2020 at 11:06:41AM -0500, Rob Landley wrote:
> On 4/20/20 5:52 PM, Andrew Morton wrote:
> > I guess it doesn't matter much, as sh seems to be rather dead.
>
> The j-core guys are using it.
Yes. There are also Debian/SH folks using classic SH4.
Rich