Re: [PATCH v2 03/13] powerpc: Remove func_descr_t

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 





Le 15/10/2021 à 00:17, Daniel Axtens a écrit :
Christophe Leroy <christophe.leroy@xxxxxxxxxx> writes:

'func_descr_t' is redundant with 'struct ppc64_opd_entry'

So, if I understand the overall direction of the series, you're
consolidating powerpc around one single type for function descriptors,
and then you're creating a generic typedef so that generic code can
always do ((func_desc_t)x)->addr to get the address of a function out of
a function descriptor regardless of arch. (And regardless of whether the
arch uses function descriptors or not.)

An architecture not using function descriptors won't do much with ((func_desc_t *)x)->addr. This is just done to allow building stuff regardless.

I prefer something like

	if (have_function_descriptors())
		addr = (func_desc_t *)ptr)->addr;
	else
		addr = ptr;

over

	#ifdef HAVE_FUNCTION_DESCRIPTORS
		addr = (func_desc_t *)ptr)->addr;
	#else
		addr = ptr;
	#endif


So:

  - why pick ppc64_opd_entry over func_descr_t?

Good question. At the begining it was because it was in UAPI headers, and also because it was the one used in our dereference_function_descriptor().

But at the end maybe that's not the more logical choice. I need to look a bit more.


  - Why not make our struct just called func_desc_t - why have a
    ppc64_opd_entry type or a func_descr_t typedef?

Well ... you usually don't flag a struct name with _t, _t will most of the time refer to a typedef.

If I want to avoid typedef (I know they are deprecated in kernel coding stype), it means the name of the struct must be changed in every architecture and it becomes tricky and it adds more churn in them, which is what I want to avoid.

At the end we risk to end-up with a messy set of #ifdefs.

Maybe this can be done as a second step, but I would like to minimise impact in this series and focus on fixing lkdtm.



  - Should this patch wait until after you've made the generic
    func_desc_t change and move directly to that new interface? (rather
    than move from func_descr_t -> ppc64_opd_entry -> ...) Or is there a
    particular reason arch specific code should use an arch-specific
    struct or named type?

As mentioned in kernel coding style, typedefs reduce readability, see https://www.kernel.org/doc/html/latest/process/coding-style.html#typedefs

But yes, we could make a step in the direction of a common 'struct func_desc'. Let's see if I can do that.

Thanks for your comments.

Christophe


Remove it.

Signed-off-by: Christophe Leroy <christophe.leroy@xxxxxxxxxx>
---
  arch/powerpc/include/asm/code-patching.h | 2 +-
  arch/powerpc/include/asm/types.h         | 6 ------
  arch/powerpc/kernel/signal_64.c          | 8 ++++----
  3 files changed, 5 insertions(+), 11 deletions(-)

diff --git a/arch/powerpc/include/asm/code-patching.h b/arch/powerpc/include/asm/code-patching.h
index 4ba834599c4d..f3445188d319 100644
--- a/arch/powerpc/include/asm/code-patching.h
+++ b/arch/powerpc/include/asm/code-patching.h
@@ -110,7 +110,7 @@ static inline unsigned long ppc_function_entry(void *func)
  	 * function's descriptor. The first entry in the descriptor is the
  	 * address of the function text.
  	 */
-	return ((func_descr_t *)func)->entry;
+	return ((struct ppc64_opd_entry *)func)->addr;
  #else
  	return (unsigned long)func;
  #endif
diff --git a/arch/powerpc/include/asm/types.h b/arch/powerpc/include/asm/types.h
index f1630c553efe..97da77bc48c9 100644
--- a/arch/powerpc/include/asm/types.h
+++ b/arch/powerpc/include/asm/types.h
@@ -23,12 +23,6 @@
typedef __vector128 vector128; -typedef struct {
-	unsigned long entry;
-	unsigned long toc;
-	unsigned long env;
-} func_descr_t;

I was a little concerned about going from a 3-element struct to a
2-element struct (as ppc64_opd_entry doesn't have an element for env) -
but we don't seem to take the sizeof this anywhere, nor do we use env
anywhere, nor do we do funky macro stuff with it in the signal handling
code that might implictly use the 3rd element, so I guess this will
work. Still, func_descr_t seems to describe the underlying ABI better
than ppc64_opd_entry...

  #endif /* __ASSEMBLY__ */
#endif /* _ASM_POWERPC_TYPES_H */
diff --git a/arch/powerpc/kernel/signal_64.c b/arch/powerpc/kernel/signal_64.c
index 1831bba0582e..63ddbe7b108c 100644
--- a/arch/powerpc/kernel/signal_64.c
+++ b/arch/powerpc/kernel/signal_64.c
@@ -933,11 +933,11 @@ int handle_rt_signal64(struct ksignal *ksig, sigset_t *set,
  		 * descriptor is the entry address of signal and the second
  		 * entry is the TOC value we need to use.
  		 */
-		func_descr_t __user *funct_desc_ptr =
-			(func_descr_t __user *) ksig->ka.sa.sa_handler;
+		struct ppc64_opd_entry __user *funct_desc_ptr =
+			(struct ppc64_opd_entry __user *)ksig->ka.sa.sa_handler;
- err |= get_user(regs->ctr, &funct_desc_ptr->entry);
-		err |= get_user(regs->gpr[2], &funct_desc_ptr->toc);
+		err |= get_user(regs->ctr, &funct_desc_ptr->addr);
+		err |= get_user(regs->gpr[2], &funct_desc_ptr->r2);

Likewise, r2 seems like a worse name than toc. I guess we could clean
that up another time though.

Kind regards,
Daniel

  	}
/* enter the signal handler in native-endian mode */
--
2.31.1



[Index of Archives]     [Linux Kernel]     [Kernel Newbies]     [x86 Platform Driver]     [Netdev]     [Linux Wireless]     [Netfilter]     [Bugtraq]     [Linux Filesystems]     [Yosemite Discussion]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Device Mapper]

  Powered by Linux