Re: [RFC PATCH 3/5] kasan: allow architectures to provide an outline readiness check

Hi Daniel,

Le 18/02/2019 à 07:13, Daniel Axtens a écrit :

christophe leroy [off-list ref] writes:

quoted

Le 15/02/2019 à 01:04, Daniel Axtens a écrit :

quoted

In powerpc (as I understand it), we spend a lot of time in boot
running in real mode before MMU paging is initalised. During
this time we call a lot of generic code, including printk(). If
we try to access the shadow region during this time, things fail.

My attempts to move early init before the first printk have not
been successful. (Both previous RFCs for ppc64 - by 2 different
people - have needed this trick too!)

So, allow architectures to define a check_return_arch_not_ready()
hook that bails out of check_memory_region_inline() unless the
arch has done all of the init.

Link: https://lore.kernel.org/patchwork/patch/592820/ # ppc64 hash series
Link: https://patchwork.ozlabs.org/patch/795211/      # ppc radix series
Originally-by: Balbir Singh [off-list ref]
Cc: Aneesh Kumar K.V <redacted>
Signed-off-by: Daniel Axtens <redacted>
---
   include/linux/kasan.h | 4 ++++
   mm/kasan/generic.c    | 2 ++
   2 files changed, 6 insertions(+)

diff --git a/include/linux/kasan.h b/include/linux/kasan.h
index f6261840f94c..83edc5e2b6a0 100644
--- a/include/linux/kasan.h
+++ b/include/linux/kasan.h

@@ -14,6 +14,10 @@ struct task_struct;
   #include <asm/kasan.h>
   #include <asm/pgtable.h>
   
+#ifndef check_return_arch_not_ready
+#define check_return_arch_not_ready()	do { } while (0)
+#endif

A static inline would be better I believe.

Something like

#ifndef kasan_arch_is_ready
static inline bool kasan_arch_is_ready {return true;}
#endif

quoted

+
   extern unsigned char kasan_early_shadow_page[PAGE_SIZE];
   extern pte_t kasan_early_shadow_pte[PTRS_PER_PTE];
   extern pmd_t kasan_early_shadow_pmd[PTRS_PER_PMD];

diff --git a/mm/kasan/generic.c b/mm/kasan/generic.c
index bafa2f986660..4c18bbd09a20 100644
--- a/mm/kasan/generic.c
+++ b/mm/kasan/generic.c

@@ -170,6 +170,8 @@ static __always_inline void check_memory_region_inline(unsigned long addr,
   						size_t size, bool write,
   						unsigned long ret_ip)
   {
+	check_return_arch_not_ready();
+

Not good for readibility that the above macro embeds a return, something
like below would be better I think:

	if (!kasan_arch_is_ready())
		return;

Unless somebody minds, I'll do the change and take this patch in my
series in order to handle the case of book3s/32 hash.

Please do; feel free to take as many of the patches as you would like
and I'll rebase whatever is left on the next version of your series.

I have now done a big step with v7: works on both nohash and hash ppc32 
without any special feature in the core of kasan. Have to do more tests 
on the hash version, but it seems promissing.

I have kept your patches on sync on top of it (allthough totally 
untested), you can find them in 
https://github.com/chleroy/linux/commits/kasan

The idea with the macro magic was to take advantage of the speed of
static keys (I think, I borrowed it from Balbir's patch). Perhaps an
inline function will achieve this anyway, but given that KASAN with
outline instrumentation is inevitably slow, I guess it doesn't matter
much either way.

You'll see in the modifications I've done to your patches, we can still 
use static keys while using static inline functions.

Christophe