mbox series

[0/3] slub: Print non-hashed pointers in slub debugging

Message ID 20210520013539.3733631-1-swboyd@chromium.org (mailing list archive)
Headers show
Series slub: Print non-hashed pointers in slub debugging | expand

Message

Stephen Boyd May 20, 2021, 1:35 a.m. UTC
I was doing some debugging recently and noticed that my pointers were
being hashed while slub_debug was on the kernel commandline. Let's make
the prints in here meaningful in that case by pushing %px throughout.

Alternatively, we could force on no_hash_pointers if slub_debug is on
the commandline. Maybe that would be better?

The final patch is just something else I noticed while looking at the
code. The message argument is never used so the debugging messages are
not as clear as they could be.

Stephen Boyd (3):
  lib/hexdump: Add a raw pointer printing format for slub debugging
  slub: Print raw pointer addresses when debugging
  slub: Actually use 'message' in restore_bytes()

 include/linux/printk.h |  1 +
 lib/hexdump.c          | 12 ++++++++++--
 mm/slub.c              | 24 ++++++++++++------------
 3 files changed, 23 insertions(+), 14 deletions(-)


base-commit: 6efb943b8616ec53a5e444193dccf1af9ad627b5

Comments

Vlastimil Babka May 20, 2021, 10:01 a.m. UTC | #1
+CC folks from the no_hash_pointers debate

Full thread with patches here:
https://lore.kernel.org/linux-mm/20210520013539.3733631-1-swboyd@chromium.org/

On 5/20/21 3:35 AM, Stephen Boyd wrote:
> I was doing some debugging recently and noticed that my pointers were
> being hashed while slub_debug was on the kernel commandline. Let's make
> the prints in here meaningful in that case by pushing %px throughout.

But we actually added no_hash_pointers exactly so that we don't push %px in
hexdump and others. So I don't think this will be accepted.

> Alternatively, we could force on no_hash_pointers if slub_debug is on
> the commandline. Maybe that would be better?

That would certainly be more acceptable, but maybe still too much, dunno. I'm
neutral on this approach, let's see what others think.

> The final patch is just something else I noticed while looking at the
> code. The message argument is never used so the debugging messages are
> not as clear as they could be.
> 
> Stephen Boyd (3):
>   lib/hexdump: Add a raw pointer printing format for slub debugging
>   slub: Print raw pointer addresses when debugging
>   slub: Actually use 'message' in restore_bytes()
> 
>  include/linux/printk.h |  1 +
>  lib/hexdump.c          | 12 ++++++++++--
>  mm/slub.c              | 24 ++++++++++++------------
>  3 files changed, 23 insertions(+), 14 deletions(-)
> 
> 
> base-commit: 6efb943b8616ec53a5e444193dccf1af9ad627b5
>