From: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
To: Christoph Hellwig <hch@infradead.org>,
linux-kernel@vger.kernel.org, shemminger@osdl.org
Subject: Re: [PATCH 1/3] Kprobes: Make kprobe modules more portable
Date: Wed, 9 Aug 2006 15:21:11 +0530 [thread overview]
Message-ID: <20060809095111.GC20050@in.ibm.com> (raw)
In-Reply-To: <20060808162421.GA28647@infradead.org>
On Tue, Aug 08, 2006 at 05:24:21PM +0100, Christoph Hellwig wrote:
> On Mon, Aug 07, 2006 at 05:25:37PM +0530, Ananth N Mavinakayanahalli wrote:
> > From: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
> >
> > This patch introduces KPROBE_ADDR, a macro that abstracts out the
> > architecture-specific artefacts of getting the correct text address
> > given a symbol. While we are at it, also introduce the symbol_name field
> > in struct kprobe to allow for users to just specify the address to be
> > probed in terms of the kernel symbol. In-kernel kprobes infrastructure
> > decodes the actual text address to probe. The symbol resolution happens
> > only if the kprobe.addr isn't explicitly specified.
>
> This looks good. A few issues are left:
>
> - the KPROBE_ADDR macro is all uppercase and not exactly very descriptive.
> - the symbol name variant should be the default, and no one outside
> kprobes.c should know about the KPROBE_ADDR macro
> - we should return EINVAL instead of silently discarding things if people
> specify a symbol and an address.
> - we should have and offset into the symbol specified
Agreed.
> The updated patch below does that, aswell as updating the only inkernel
> kprobes user (tcp_probe.c) to the new interface (*) and removing the now
> obsolete kallsysms_lookup_name export.
>
> (*) tcp_probe.c shows very well how horrible the old interface was, as it's
> not portable to ppc64 as-is
>
> Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
> Index: linux-2.6/include/asm-powerpc/kprobes.h
> ===================================================================
> --- linux-2.6.orig/include/asm-powerpc/kprobes.h 2006-08-08 17:47:22.000000000 +0200
> +++ linux-2.6/include/asm-powerpc/kprobes.h 2006-08-08 18:13:57.000000000 +0200
> @@ -44,6 +44,9 @@
> #define IS_TDI(instr) (((instr) & 0xfc000000) == 0x08000000)
> #define IS_TWI(instr) (((instr) & 0xfc000000) == 0x0c000000)
>
> +#define kprobe_lookup_name(name) \
> + (*((kprobe_opcode_t **)kallsyms_lookup_name(name)))
> +
> #define JPROBE_ENTRY(pentry) (kprobe_opcode_t *)((func_descr_t *)pentry)
>
> #define is_trap(instr) (IS_TW(instr) || IS_TD(instr) || \
> Index: linux-2.6/include/linux/kprobes.h
> ===================================================================
> --- linux-2.6.orig/include/linux/kprobes.h 2006-08-08 17:47:22.000000000 +0200
> +++ linux-2.6/include/linux/kprobes.h 2006-08-08 17:47:31.000000000 +0200
> @@ -77,6 +77,12 @@
> /* location of the probe point */
> kprobe_opcode_t *addr;
>
> + /* Allow user to indicate symbol name of the probe point */
> + char *symbol_name;
> +
> + /* Offset into the symbol */
> + unsigned int offset;
> +
> /* Called before addr is executed. */
> kprobe_pre_handler_t pre_handler;
>
> Index: linux-2.6/kernel/kprobes.c
> ===================================================================
> --- linux-2.6.orig/kernel/kprobes.c 2006-08-08 17:47:22.000000000 +0200
> +++ linux-2.6/kernel/kprobes.c 2006-08-08 17:47:31.000000000 +0200
> @@ -37,6 +37,7 @@
> #include <linux/slab.h>
> #include <linux/module.h>
> #include <linux/moduleloader.h>
> +#include <linux/kallsyms.h>
> #include <asm-generic/sections.h>
> #include <asm/cacheflush.h>
> #include <asm/errno.h>
> @@ -45,6 +46,16 @@
> #define KPROBE_HASH_BITS 6
> #define KPROBE_TABLE_SIZE (1 << KPROBE_HASH_BITS)
>
> +
> +/*
> + * Some oddball architectures like 64bit powerpc have function descriptors
> + * so this must be overridable.
> + */
> +#ifndef kprobe_lookup_name
> +#define kprobe_lookup_name(name) \
> + ((kprobe_opcode_t *)(kallsyms_lookup_name(name))
> +#endif
> +
> static struct hlist_head kprobe_table[KPROBE_TABLE_SIZE];
> static struct hlist_head kretprobe_inst_table[KPROBE_TABLE_SIZE];
> static atomic_t kprobe_count;
> @@ -447,6 +458,17 @@
> struct kprobe *old_p;
> struct module *probed_mod;
>
> + /*
> + * If we have a symbol_name argument look it up,
> + * and add it to the address. That way the addr
> + * field can either be global or relative to a symbol.
> + */
> + if (p->symbol_name) {
> + if (p->addr)
> + return -EINVAL;
> + p->addr = kprobe_lookup_name(p->symbol_name) + p->offset;
> + }
> +
> if ((!kernel_text_address((unsigned long) p->addr)) ||
> in_kprobes_functions((unsigned long) p->addr))
> return -EINVAL;
> Index: linux-2.6/kernel/kallsyms.c
> ===================================================================
> --- linux-2.6.orig/kernel/kallsyms.c 2006-08-08 17:13:14.000000000 +0200
> +++ linux-2.6/kernel/kallsyms.c 2006-08-08 17:47:39.000000000 +0200
> @@ -154,7 +154,6 @@
> }
> return module_kallsyms_lookup_name(name);
> }
> -EXPORT_SYMBOL_GPL(kallsyms_lookup_name);
>
> /*
> * Lookup an address
> Index: linux-2.6/net/ipv4/tcp_probe.c
> ===================================================================
> --- linux-2.6.orig/net/ipv4/tcp_probe.c 2006-08-08 18:13:55.000000000 +0200
> +++ linux-2.6/net/ipv4/tcp_probe.c 2006-08-08 18:14:28.000000000 +0200
> @@ -99,8 +99,10 @@
> }
>
> static struct jprobe tcp_send_probe = {
> - .kp = { .addr = (kprobe_opcode_t *) &tcp_sendmsg, },
> - .entry = (kprobe_opcode_t *) &jtcp_sendmsg,
> + .kp = {
> + .symbol_name = "tcp_sendmsg",
> + },
> + .entry = JPROBE_ENTRY(jtcp_sendmsg),
> };
>
>
next prev parent reply other threads:[~2006-08-09 9:49 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-08-07 11:55 [PATCH 1/3] Kprobes: Make kprobe modules more portable Ananth N Mavinakayanahalli
2006-08-07 12:00 ` [PATCH 2/3] Kprobes: Define retval helper Ananth N Mavinakayanahalli
2006-08-07 12:04 ` [PATCH 3/3] Kprobes: Update Documentation/kprobes.txt Ananth N Mavinakayanahalli
2006-08-08 16:27 ` Christoph Hellwig
2006-08-09 9:48 ` Ananth N Mavinakayanahalli
2006-08-09 10:24 ` Ananth N Mavinakayanahalli
2006-08-08 16:25 ` [PATCH 2/3] Kprobes: Define retval helper Christoph Hellwig
2006-08-09 4:37 ` Andi Kleen
2006-08-09 9:43 ` Ananth N Mavinakayanahalli
2006-08-09 9:45 ` Christoph Hellwig
2006-08-09 9:55 ` Andi Kleen
2006-08-09 10:19 ` Ananth N Mavinakayanahalli
2006-08-09 10:28 ` Martin Schwidefsky
2006-08-09 13:16 ` David Howells
2006-08-09 15:20 ` Ananth N Mavinakayanahalli
2006-08-07 12:06 ` [PATCH 1/3] Kprobes: Make kprobe modules more portable Ananth N Mavinakayanahalli
2006-08-08 16:24 ` Christoph Hellwig
2006-08-08 16:34 ` Stephen Hemminger
2006-08-08 16:40 ` Christoph Hellwig
2006-08-08 17:28 ` Paulo Marques
2006-08-08 17:32 ` Christoph Hellwig
2006-08-09 9:51 ` Ananth N Mavinakayanahalli [this message]
2006-08-09 16:01 ` David Smith
2006-08-09 16:10 ` Christoph Hellwig
2006-08-09 16:18 ` Christoph Hellwig
2006-08-10 14:10 ` Keshavamurthy Anil S
2006-08-16 13:20 ` Christoph Hellwig
2006-08-11 8:53 ` Ananth N Mavinakayanahalli
2006-08-10 6:18 ` Ananth N Mavinakayanahalli
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=20060809095111.GC20050@in.ibm.com \
--to=ananth@in.ibm.com \
--cc=hch@infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=shemminger@osdl.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.