All of lore.kernel.org
 help / color / mirror / Atom feed
From: Tony Wu <tung7970@gmail.com>
To: Ralf Baechle <ralf@linux-mips.org>
Cc: Alan Cooper <alcooperx@gmail.com>,
	Petri Gynther <pgynther@google.com>,
	Jun-Ru Chang <jrjang@gmail.com>,
	cminyard@mvista.com, Steven Rostedt <rostedt@goodmis.org>,
	linux-mips@linux-mips.org
Subject: Re: [PATCH] MIPS: Ftrace: Fix dynamic tracing of kernel modules
Date: Fri, 8 Aug 2014 12:11:54 +0800	[thread overview]
Message-ID: <20140808120113-tung7970@googlemail.com> (raw)
In-Reply-To: <20140807212514.GD29898@linux-mips.org>

On Thu, Aug 07, 2014 at 11:25:14PM +0200, Ralf Baechle wrote:
> On Wed, Aug 06, 2014 at 01:12:06PM -0400, Alan Cooper wrote:
> 
> > Actually , there's no reason to write the second NOP when nop'ing the
> > mcount call site in a module. This was done to remove the stack adjust
> > instruction which only exists at this location for internal kernel
> > routines. The following diff seems like a simpler way to solve issue
> > #1:
> 
> Oh?
> 
> $ mips-linux-objdump -d --reloc net/sctp/sctp.ko
> [...]
> 00000000 <sctp_sm_lookup_event>:
>        0:       27bdffe8        addiu   sp,sp,-24
>        4:       afbf0014        sw      ra,20(sp)
>        8:       3c030000        lui     v1,0x0
>                         8: R_MIPS_HI16  _mcount
>        c:       24630000        addiu   v1,v1,0
>                         c: R_MIPS_LO16  _mcount
>       10:       03e00821        move    at,ra
>       14:       27ac0014        addiu   t4,sp,20
>       18:       0060f809        jalr    v1
>       1c:       27bdfff8        addiu   sp,sp,-8  <====
> [...]
>       64:       27bd0018        addiu   sp,sp,24
>       68:       03e00008        jr      ra
> [...]
> 
> So the stack adjustment also exists for modules.
> 
> Or am I missunderstanding something?
> 
>   Ralf
> 

We have a workaround for dynamic ftrace under 32bit mode back in Feburary. I thought
it is not good enough but maybe this is the right solution for issue #1 ?

Tony

commit d3167328b1bd63c22abb129a17fcb658e11c2a7b
Author: Jun-Ru Chang <jrjang@gmail.com>
Date:   Thu Feb 27 15:23:34 2014 +0800

    mips: ftrace: fix ftrace_make_call long call restore
    
    In case of long call under 32bit mode, two instructions,
    lui and addiu, are required to load the mcount address into
    the jump register.
    
    In ftrace_make_nop, both instructions are marked as nop, so
    in ftrace_make_call, we have to restore both of them.
    
    Signed-off-by: Jun-Ru Chang <jrjang@gmail.com>
    Signed-off-by: Tony Wu <tung7970@gmail.com>

diff --git a/arch/mips/kernel/ftrace.c b/arch/mips/kernel/ftrace.c
index 9bb8bcd..b12858c 100644
--- a/arch/mips/kernel/ftrace.c
+++ b/arch/mips/kernel/ftrace.c
@@ -61,6 +61,9 @@ static inline int in_kernel_space(unsigned long ip)
 
 static unsigned int insn_jal_ftrace_caller __read_mostly;
 static unsigned int insn_lui_v1_hi16_mcount __read_mostly;
+#ifndef CONFIG_64BIT
+static unsigned int insn_add_v1_lo16_mcount __read_mostly;
+#endif
 static unsigned int insn_j_ftrace_graph_caller __maybe_unused __read_mostly;
 
 static inline void ftrace_dyn_arch_init_insns(void)
@@ -73,6 +76,12 @@ static inline void ftrace_dyn_arch_init_insns(void)
 	buf = (u32 *)&insn_lui_v1_hi16_mcount;
 	UASM_i_LA_mostly(&buf, v1, MCOUNT_ADDR);
 
+#ifndef CONFIG_64BIT
+	/* addiu v1, vi, lo16_mcount */
+	buf = (u32 *)&insn_add_v1_lo16_mcount;
+	UASM_i_LA(&buf, v1, MCOUNT_ADDR);
+#endif
+
 	/* jal (ftrace_caller + 8), jump over the first two instruction */
 	buf = (u32 *)&insn_jal_ftrace_caller;
 	uasm_i_jal(&buf, (FTRACE_ADDR + 8) & JUMP_RANGE_MASK);
@@ -180,7 +189,10 @@ int ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 	new = in_kernel_space(ip) ? insn_jal_ftrace_caller :
 		insn_lui_v1_hi16_mcount;
 
-	return ftrace_modify_code(ip, new);
+	if (IS_BUILTIN(CONFIG_64BIT) || new == insn_jal_ftrace_caller)
+		return ftrace_modify_code(ip, new);
+	else
+		return ftrace_modify_code_2(ip, new, insn_add_v1_lo16_mcount);
 }
 
 #define FTRACE_CALL_IP ((unsigned long)(&ftrace_call))

      reply	other threads:[~2014-08-08  4:12 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-07-24  5:55 [PATCH] MIPS: Ftrace: Fix dynamic tracing of kernel modules Petri Gynther
2014-08-05 16:41 ` Alan Cooper
     [not found]   ` <CAGXr9JE7v9-hS3irmdgeaEU2iGLZHshEr_N-Do1UAsZhyzMe2g@mail.gmail.com>
2014-08-06 17:12     ` Alan Cooper
2014-08-07 21:25       ` Ralf Baechle
2014-08-08  4:11         ` Tony Wu [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20140808120113-tung7970@googlemail.com \
    --to=tung7970@gmail.com \
    --cc=alcooperx@gmail.com \
    --cc=cminyard@mvista.com \
    --cc=jrjang@gmail.com \
    --cc=linux-mips@linux-mips.org \
    --cc=pgynther@google.com \
    --cc=ralf@linux-mips.org \
    --cc=rostedt@goodmis.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.