From: Hui Zhu <teawater@gmail.com>
To: Andi Kleen <andi@firstfloor.org>
Cc: linux-kernel@vger.kernel.org,
Christoph Hellwig <hch@infradead.org>,
Geoff Levand <geoff@infradead.org>,
jason.wessel@windriver.com
Subject: Re: [PATCH]KGTP (Linux Kernel debugger and tracer) lite patch for review
Date: Thu, 24 May 2012 21:35:52 +0800 [thread overview]
Message-ID: <4FBE3938.30000@gmail.com> (raw)
In-Reply-To: <20120510173808.GA27374@one.firstfloor.org>
On Fri, May 11, 2012 at 1:38 AM, Andi Kleen <andi@firstfloor.org> wrote:
> On Thu, May 10, 2012 at 08:15:36PM +0800, Hui Zhu wrote:
>> On 05/09/12 22:05, Andi Kleen wrote:
>> >Please provide better explanation of the use case for this module.
>> >One paragraph why someone would want it in their kernel.
>>
>> sudo insmod kernel/gtp.ko
>
> Thanks. Please put that into the change log for the next iteration.
OK. I will add them.
>
>> >>+#ifndef _ASM_ARM_GTP_H_
>> >>+#define _ASM_ARM_GTP_H_
>> >>+
>> >>+#define ULONGEST uint64_t
>> >
>> >So u64 in kernel. Just use that.
>> >
>> >>+#define CORE_ADDR unsigned long
>> >
>> >In linux kernel CORE_ADDR is always unsigned long. Use that.
>> >
>>
>> ULONGEST and CORE_ADDR is the type from GDB rsp packet. I use it to parse
>> the package from GDB. Then want to add add a new arch to KGTP, just need
>> set this two type same with GDB. So do you mind I keep it?
>
> I was just aiming to eliminate the asm files. Is that possible?
> If you can't please reuse stuff from kgdb.
I remove them from asm files, and add:
#ifndef ULONGEST
#define ULONGEST uint64_t
#endif
#ifndef CORE_ADDR
#define CORE_ADDR unsigned long
#endif
in gtp.c
Then if not need, the arch didn't define ULONGEST and CORE_ADDR for itself.
>
>> OK. I changed hex2ulongest to simple_strtoull in new patch.
>> But I meet a issue is checkpatch told me that "simple_strtoull is
>> obsolete, use kstrtoull instead". But kstrtoull cannot get the address
>> that where this convert stop at. Do you have some comments on it?
>
> Keep using simple_strtoull. Ignore checkpatch when it's wrong.
>
>> +static inline void
>> +gtp_regs2ascii(struct pt_regs *regs, char *buf)
>> +{
>> +#ifdef CONFIG_32BIT
>
> BITS_PER_LONG == 32 ? Then it would work on other archs and you could
> move it out from asm/
That is hard becuase the different arch's format of reg is so different.
For example ARM and MIPS have a lot of special reg that cannot be converted by a for loop.
After check the code the kgdb, I thought that good ways is use dbg_reg_def of kgdb to convert the reg to gdb rsp format. But use it directly will make kgtp depend on kgdb.
What I thought is move this part of code out from kgdb, when kgdb or kgtp need, select it too. But I am not sure Jason is OK with that.
>
>> +#define OUTFORMAT "%08lx"
>> +#define REGSIZE 8
>> +#define SWAB(a) ntohl(a)
>> +#else
>> +#define OUTFORMAT "%016lx"
>> +#define REGSIZE 16
>> +#ifdef __LITTLE_ENDIAN
>> +#define SWAB(a) swab64(a)
>> +#else
>> +#define SWAB(a) (a)
>> +#endif
>
> be64_to_cpu()
OK.
>
>> @@ -0,0 +1,914 @@
>> +/*
>> + * Kernel GDB tracepoint module.
>
> add one sentence here what this file does.
OK.
>
>> + memcpy(&freg->regs, regs, sizeof(struct pt_regs));
>> +#ifdef CONFIG_X86_32
>> + freg->regs.sp = (unsigned long)®s->sp;
>> +#endif /* CONFIG_X86_32 */
>> +#ifdef CONFIG_X86
>> + freg->regs.ip -= 1;
>> +#endif /* CONFIG_X86 */
>
> What is that for? - 1?
That is because when kprobe, the ip of X86 point will have a offset with current ip.
>
>> +
>> +static void
>> +gtp_action_release(struct action *ae)
>
> Standard style is type on the same line.
OK. Fixed them all in the patch.
>
>> +{
>> + struct action *ae2;
>> +
>> + while (ae) {
>> + ae2 = ae;
>> + ae = ae->next;
>> + /* Release ae2. */
>> + kfree(ae2);
>> + }
>
> Better use a list_head from list.h and list_for_each_entry_safe() etc.
> This will simplify some of the list code.
OK. I changed both gtp_list and action to list_head.
>
>> +static int
>> +gtp_gdbrsp_qtstart(void)
>> +{
>> + struct gtp_entry *tpe;
>> +
>> + if (gtp_start)
>> + return -EBUSY;
>
> This is in a lot of places. Can you put it somewhere more central?
Sorry I don't understand this part. Do you want I change all "struct gtp_entry *tpe;" to "struct gtp_entry *tpe;"?
>> + if (!gtp_frame) {
>> + gtp_frame = vmalloc(GTP_FRAME_SIZE);
>> + if (!gtp_frame)
>> + return -ENOMEM;
>> +
>> + gtp_frame_reset();
>> + }
>> +
>> + for (tpe = gtp_list; tpe; tpe = tpe->next) {
>> + if (tpe->action_list) {
>> + int ret = register_kprobe(&tpe->kp);
>> + if (ret < 0) {
>> + gtp_gdbrsp_qtstop();
>> + return ret;
>
> memory leak with frame?
Use this because I don't want alloc frame each time tracepoint start.
It will be release in gtp_release.
>
>> + pr_devel("gtp_gdbrsp_QT: %s\n", pkg);
>> +
>> + if (strcmp("init", pkg) == 0)
>> + ret = gtp_gdbrsp_qtinit();
>
> This if cascade should be probably a table walk
I cannot find the examle about it in the Kernel, could you give me a example?
>
>> + else if (strcmp("Stop", pkg) == 0)
>> + ret = gtp_gdbrsp_qtstop();
>> +
>> + return ret;
>> +}
>> +
>> +static unsigned char gtp_m_buffer[0xffff];
>
> Ok so what lock protects this buffer?
No. The gtp_frame_lock protects the vars that follow it:
static int gtp_frame_num;
static char *gtp_frame;
static char *gtp_frame_r_start;
static char *gtp_frame_w_start;
static char *gtp_frame_w_end;
static char *gtp_frame_r_cache;
static int gtp_frame_is_circular;
It protects them all.
gtp_m_buffer is protected by gtp_rw_lock that I just add in new patch.
>
>> + struct gtp_frame_reg *fr = NULL;
>> +
>> + if (GTP_RW_MAX - 4 - gtp_rw_size < GTP_REG_ASCII_SIZE)
>> + return -E2BIG;
>
> The magic 4 should be a define
Done.
>
>> + next = gtp_frame_current->next;
>> + while (next) {
>> + switch (next[0]) {
>> + case 'r':
>> + fr = (struct gtp_frame_reg *) (next + 1);
>> + goto check;
>> + break;
>
> either check or break, but not both
>
>> + default:
>> + next = NULL;
>
> Not needed
>> + break;
>> + }
>> + }
This part is fixed.
>> +check:
>> + if (fr)
>> + gtp_regs2ascii(&fr->regs, gtp_rw_bufp);
>> + else
>> + memset(gtp_rw_bufp, '0', GTP_REG_ASCII_SIZE);
>
> No 0 termination?
No, it don't need becuase it will add head and tail in gtp_write.
>
>> +
>> +static int
>> +gtp_open(struct inode *inode, struct file *file)
>> +{
>> + if (atomic_inc_return(>p_count) > 1) {
>> + atomic_dec(>p_count);
>> + return -EBUSY;
>> + }
>
> This looks racy. Better use a try mutex lock
Fixed.
>
>> +static int __init gtp_init(void)
>> +{
>> + atomic_set(>p_count, 0);
>> + gtp_read_ack = 0;
>> + gtp_rw_buf = NULL;
>> + gtp_rw_bufp = NULL;
>> + gtp_rw_size = 0;
>> + gtp_start = 0;
>> + gtp_disconnected_tracing = 0;
>> + gtp_circular = 0;
>> + gtp_frame_num = 0;
>> + gtp_frame = NULL;
>> + gtp_frame_r_start = NULL;
>> + gtp_frame_w_start = NULL;
>> + gtp_frame_w_end = NULL;
>> + gtp_frame_r_cache = NULL;
>> + gtp_frame_is_circular = 0;
>> + gtp_frame_current = NULL;
>
> Globals don't need to be initialized with 0
OK. Fixed.
>
>>
>> +config GTP
>> + tristate "GDB tracepoint support"
>> + depends on X86 || ARM || MIPS
>> + select KPROBES
>> + select DEBUG_FS
>> + ---help---
>> + Supply GDB tracepoint interface in /sys/kernel/debug/gtp.
>
> Need far more description here. Write 1-2 paragraphs why a user
> wants this. checkpatch should have warned.
OK. Add some introduce. checkpatch is OK with it.
>
> -Andi
> --
> ak@linux.intel.com -- Speaking for myself only.
Thanks for your help.
Best,
Hui
sudo insmod kernel/gtp.ko
sudo gdb ./vmlinux
#Connect KGTP interface
(gdb) target remote /sys/kernel/debug/gtp
Remote debugging using /sys/kernel/debug/gtp
0x0000000000000000 in irq_stack_union ()
#Access memory directly
(gdb) p jiffies_64
$1 = 4332444293
(gdb) p jiffies_64
$2 = 4332444591
(gdb) p *((struct module *)((char *)modules->next - ((size_t) &(((struct module *)0)->list))))
$3 = {state = MODULE_STATE_LIVE, list = {next = 0xffffffffa0964188, prev = 0xffffffff81c345f0},
name = "gtp", '\000' <repeats 52 times>, mkobj = {kobj = {name = 0xffff88011f5c49c8 "gtp", entry = {
next = 0xffff880221cd6420, prev = 0xffffffffa09641d8}, parent = 0xffff880221cd6438, kset = 0xffff880221cd6420,
...
(gdb) p *(struct device *)(__per_cpu_offset[0]+(uint64_t)(&mce_device))
$4 = {parent = 0xffff8802212f8000, p = 0x1fe, kobj = {name = 0x0, entry = {next = 0x0, prev = 0xffffffff8102ae70},
parent = 0x0, kset = 0x0, ktype = 0x0, sd = 0x0, kref = {refcount = {counter = 0}}, state_initialized = 0,
...
#Set tracepoint to collect the register info of function vfs_read
(gdb) trace vfs_read
Tracepoint 1 at 0xffffffff8117a3d0: file /home/teawater/kernel2/linux/fs/read_write.c, line 365.
(gdb) actions
Enter actions for tracepoint 1, one per line.
End with a line saying just "end".
collect $reg
end
(gdb) tstart
(gdb) tstop
(gdb) tfind
Found trace frame 0, tracepoint 1
#0 vfs_read (file=0xffff880070bb6100,
buf=0x31cb8f0 "$2aa6178136313733#331bb700088ffff481fe4450188ffff002", '0' <repeats 13 times>, "f0b81c03000000000061bb700088ffff781fe4450188ffff301fe4450188ffff606944030000000058f19a010288ffff01", '0' <repeats 14 times>, "4602", '0' <repeats 12 times>, "f0b81c0"..., count=8192, pos=0xffff880145e41f48) at /home/teawater/kernel2/linux/fs/read_write.c:365
365 {
(gdb) info reg
rax 0x0 0
rbx 0xffff880070bb6100 -131939504004864
rcx 0xffff880145e41f48 -131935927787704
rdx 0x2000 8192
rsi 0x31cb8f0 52214000
rdi 0xffff880070bb6100 -131939504004864
rbp 0xffff880145e41f78 0xffff880145e41f78
rsp 0xffff880145e41f30 0xffff880145e41f30
r8 0x3446960 54815072
r9 0xffff8802019af158 -131932778466984
r10 0x1 1
r11 0x246 582
r12 0x31cb8f0 52214000
r13 0x2000 8192
r14 0xa 10
r15 0xa 10
rip 0xffffffff8117a3d0 0xffffffff8117a3d0 <vfs_read>
eflags 0x286 [ PF SF IF ]
cs 0x10 16
ss 0x18 24
ds *value not available*
es *value not available*
fs *value not available*
gs *value not available*
Because this is a lite patch, so this patch doesn't support other functions of KGTP.
Signed-off-by: Hui Zhu <teawater@gmail.com>
---
arch/arm/include/asm/gtp.h | 24 +
arch/mips/include/asm/gtp.h | 56 ++
arch/x86/include/asm/gtp.h | 92 ++++
kernel/Makefile | 2
kernel/gtp.c | 892 ++++++++++++++++++++++++++++++++++++++++++++
lib/Kconfig.debug | 12
6 files changed, 1078 insertions(+)
--- /dev/null
+++ b/arch/arm/include/asm/gtp.h
@@ -0,0 +1,24 @@
+#ifndef _ASM_ARM_GTP_H_
+#define _ASM_ARM_GTP_H_
+
+#define GTP_REG_ASCII_SIZE 336
+
+static inline void gtp_regs2ascii(struct pt_regs *regs, char *buf)
+{
+ int i;
+
+ for (i = 0; i < 16; i++) {
+ sprintf(buf, "%08lx", (unsigned long) ntohl(regs->uregs[i]));
+ buf += 8;
+ }
+
+ /* f0-f7 fps */
+ memset(buf, '0', 200);
+ buf += 200;
+
+ sprintf(buf, "%08lx",
+ (unsigned long) ntohl(regs->uregs[16]));
+ buf += 8;
+}
+
+#endif
--- /dev/null
+++ b/arch/mips/include/asm/gtp.h
@@ -0,0 +1,56 @@
+#ifndef _ASM_MIPS_GTP_H_
+#define _ASM_MIPS_GTP_H_
+
+#ifdef CONFIG_32BIT
+#define GTP_REG_ASCII_SIZE 304
+#else
+#define GTP_REG_ASCII_SIZE 608
+#endif
+
+#define GTP_SP_NUM 29
+
+static inline void gtp_regs2ascii(struct pt_regs *regs, char *buf)
+{
+#ifdef CONFIG_32BIT
+#define OUTFORMAT "%08lx"
+#define REGSIZE 8
+#define SWAB(a) ntohl(a)
+#else
+#define OUTFORMAT "%016lx"
+#define REGSIZE 16
+#define SWAB(a) be64_to_cpu(a)
+#endif
+ {
+ int i;
+
+ for (i = 0; i < 32; i++) {
+ sprintf(buf, OUTFORMAT,
+ (unsigned long) SWAB(regs->regs[i]));
+ buf += REGSIZE;
+ }
+ }
+
+ sprintf(buf, OUTFORMAT,
+ (unsigned long) SWAB(regs->cp0_status));
+ buf += REGSIZE;
+ sprintf(buf, OUTFORMAT,
+ (unsigned long) SWAB(regs->lo));
+ buf += REGSIZE;
+ sprintf(buf, OUTFORMAT,
+ (unsigned long) SWAB(regs->hi));
+ buf += REGSIZE;
+ sprintf(buf, OUTFORMAT,
+ (unsigned long) SWAB(regs->cp0_badvaddr));
+ buf += REGSIZE;
+ sprintf(buf, OUTFORMAT,
+ (unsigned long) SWAB(regs->cp0_cause));
+ buf += REGSIZE;
+ sprintf(buf, OUTFORMAT,
+ (unsigned long) SWAB(regs->cp0_epc));
+ buf += REGSIZE;
+#undef OUTFORMAT
+#undef REGSIZE
+#undef SWAB
+}
+
+#endif
--- /dev/null
+++ b/arch/x86/include/asm/gtp.h
@@ -0,0 +1,92 @@
+#ifndef _ASM_X86_GTP_H_
+#define _ASM_X86_GTP_H_
+
+#ifdef CONFIG_X86_32
+#define GTP_REG_ASCII_SIZE 128
+#else
+#define GTP_REG_ASCII_SIZE 296
+#endif
+
+static inline void gtp_regs2ascii(struct pt_regs *regs, char *buf)
+{
+#ifdef CONFIG_X86_32
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->ax));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->cx));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->dx));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->bx));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->sp));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->bp));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->si));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->di));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->ip));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->flags));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->cs));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->ss));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->ds));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->es));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->fs));
+ buf += 8;
+ sprintf(buf, "%08x", (unsigned int) swab32(regs->gs));
+ buf += 8;
+#else
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->ax));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->bx));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->cx));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->dx));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->si));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->di));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->bp));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->sp));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->r8));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->r9));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->r10));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->r11));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->r12));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->r13));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->r14));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->r15));
+ buf += 16;
+ sprintf(buf, "%016lx", (unsigned long) swab64(regs->ip));
+ buf += 16;
+ sprintf(buf, "%08x",
+ (unsigned int) swab32((unsigned int)regs->flags));
+ buf += 8;
+ sprintf(buf, "%08x",
+ (unsigned int) swab32((unsigned int)regs->cs));
+ buf += 8;
+ sprintf(buf, "%08x",
+ (unsigned int) swab32((unsigned int)regs->ss));
+ buf += 8;
+#endif
+}
+
+#endif
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -108,6 +108,8 @@ obj-$(CONFIG_PADATA) += padata.o
obj-$(CONFIG_CRASH_DUMP) += crash_dump.o
obj-$(CONFIG_JUMP_LABEL) += jump_label.o
+obj-$(CONFIG_GTP) += gtp.o
+
$(obj)/configs.o: $(obj)/config_data.h
# config_data.h contains the same information as ikconfig.h but gzipped.
--- /dev/null
+++ b/kernel/gtp.c
@@ -0,0 +1,892 @@
+/*
+ * KGTP is a realtime and lightweight Linux debugger and tracer.
+ * It makes Linux Kernel supply a GDB remote debug interface.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * Copyright(C) KGTP team (https://code.google.com/p/kgtp/), 2010, 2011, 2012
+ *
+ */
+
+#include <linux/version.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/uaccess.h>
+#include <linux/vmalloc.h>
+#include <linux/poll.h>
+#include <linux/kprobes.h>
+#include <linux/interrupt.h>
+#include <linux/debugfs.h>
+#include <linux/slab.h>
+#include <asm/gtp.h>
+
+#ifndef ULONGEST
+#define ULONGEST uint64_t
+#endif
+#ifndef CORE_ADDR
+#define CORE_ADDR unsigned long
+#endif
+
+#define GTP_RW_MAX 16384
+#define GTP_RW_BUFP_MAX (GTP_RW_MAX - 4 - gtp_rw_size)
+
+#define GTP_FRAME_SIZE 5242880
+#define GTP_FRAME_HEAD_SIZE (1 + sizeof(struct gtp_frame_head))
+#define GTP_FRAME_REG_SIZE (1 + sizeof(struct gtp_frame_reg))
+
+#define TOHEX(h) ((h) > 9 ? (h) + 'a' - 10 : (h) + '0')
+
+struct action {
+ struct list_head node;
+ char type;
+ union {
+ ULONGEST reg_mask;
+ } u;
+};
+
+struct gtp_entry {
+ struct list_head node;
+ ULONGEST num;
+ ULONGEST addr;
+ ULONGEST step;
+ ULONGEST pass;
+ int nopass;
+ int kpreg;
+ struct kprobe kp;
+ struct list_head action_list;
+};
+
+static LIST_HEAD(gtp_list);
+
+struct gtp_frame_head {
+ int frame_num;
+ ULONGEST trace_num;
+ char *next;
+};
+
+struct gtp_frame_reg {
+ struct pt_regs regs;
+ char *next;
+};
+
+static char gtp_read_ack;
+static char *gtp_rw_buf;
+static char *gtp_rw_bufp;
+static size_t gtp_rw_size;
+
+static int gtp_start;
+
+static int gtp_disconnected_tracing;
+static int gtp_circular;
+
+/* This spin_lock protects the gtp_frame. */
+static DEFINE_SPINLOCK(gtp_frame_lock);
+static int gtp_frame_num;
+static char *gtp_frame;
+static char *gtp_frame_r_start;
+static char *gtp_frame_w_start;
+static char *gtp_frame_w_end;
+static char *gtp_frame_r_cache;
+static int gtp_frame_is_circular;
+static struct gtp_frame_head *gtp_frame_current;
+
+static DEFINE_SEMAPHORE(gtp_rw_lock);
+static int gtp_rw_count;
+
+static char *gtp_frame_alloc(size_t size)
+{
+ char *ret = NULL;
+
+ if (size > GTP_FRAME_SIZE)
+ return NULL;
+
+ spin_lock(>p_frame_lock);
+
+ if (gtp_frame_w_start + size > gtp_frame_w_end) {
+ if (gtp_circular) {
+ gtp_frame_is_circular = 1;
+ gtp_frame_w_start = gtp_frame;
+ gtp_frame_r_start = gtp_frame;
+ } else
+ goto out;
+ }
+
+ if (gtp_frame_is_circular) {
+ /* Release some frame entry to get some place.
+ XXX: When support new frame type, need add new handler
+ to switch. */
+ while (gtp_frame_w_start + size > gtp_frame_r_start) {
+ switch (gtp_frame_r_start[0]) {
+ case 'h':
+ gtp_frame_r_start += GTP_FRAME_HEAD_SIZE;
+ break;
+ case 'r':
+ gtp_frame_r_start += GTP_FRAME_REG_SIZE;
+ break;
+ default:
+ goto out;
+ }
+ }
+ }
+
+ ret = gtp_frame_w_start;
+ gtp_frame_w_start += size;
+
+out:
+ spin_unlock(>p_frame_lock);
+ return ret;
+}
+
+static inline char **gtp_action_r(struct pt_regs *regs, struct action *ae,
+ char **next)
+{
+ struct gtp_frame_reg *freg;
+ char *tmp;
+
+ tmp = gtp_frame_alloc(GTP_FRAME_REG_SIZE);
+ if (!tmp)
+ return NULL;
+
+ *next = tmp;
+ tmp[0] = 'r';
+ freg = (struct gtp_frame_reg *) (tmp + 1);
+ memcpy(&freg->regs, regs, sizeof(struct pt_regs));
+#ifdef CONFIG_X86_32
+ freg->regs.sp = (unsigned long)®s->sp;
+#endif /* CONFIG_X86_32 */
+#ifdef CONFIG_X86
+ freg->regs.ip -= 1;
+#endif /* CONFIG_X86 */
+ freg->next = NULL;
+
+ return &freg->next;
+}
+
+static int gtp_kp_pre_handler(struct kprobe *p, struct pt_regs *regs)
+{
+ struct gtp_entry *tpe = container_of(p, struct gtp_entry, kp);
+ struct gtp_frame_head *head;
+ char *tmp;
+ struct action *ae;
+ struct list_head *cur;
+ char **next;
+
+ pr_devel("gtp_kp_pre_handler: tracepoint %d\n", (int)tpe->num);
+
+ /* Get the head. */
+ tmp = gtp_frame_alloc(GTP_FRAME_HEAD_SIZE);
+ if (!tmp)
+ goto no_memory;
+ tmp[0] = 'h';
+ head = (struct gtp_frame_head *) (tmp + 1);
+ /* Get a new frame num from gtp_frame_num. */
+ spin_lock(>p_frame_lock);
+ if (gtp_frame_num < 0)
+ gtp_frame_num = head->frame_num = 0;
+ else
+ head->frame_num = gtp_frame_num++;
+ spin_unlock(>p_frame_lock);
+ head->trace_num = tpe->num;
+ head->next = NULL;
+ next = &head->next;
+
+ /* Handle actions. */
+ list_for_each(cur, &tpe->action_list) {
+ ae = list_entry(cur, struct action, node);
+ switch (ae->type) {
+ case 'r':
+ next = gtp_action_r(regs, ae, next);
+ if (!next)
+ goto no_memory;
+ break;
+ }
+ }
+
+ return 0;
+
+no_memory:
+ pr_devel("gtp_kp_pre_handler: tracepoint %d no memory.\n",
+ (int)tpe->num);
+ return 0;
+}
+
+static struct action *gtp_action_alloc(char type)
+{
+ struct action *ret;
+
+ ret = kzalloc(sizeof(struct action), GFP_KERNEL);
+ if (!ret)
+ goto out;
+
+ ret->type = type;
+
+out:
+ return ret;
+}
+
+static struct gtp_entry *gtp_list_add(ULONGEST num, ULONGEST addr)
+{
+ struct gtp_entry *ret = kzalloc(sizeof(struct gtp_entry),
+ GFP_KERNEL);
+
+ if (!ret)
+ goto out;
+ ret->num = num;
+ ret->addr = addr;
+ ret->kp.addr = (kprobe_opcode_t *) (CORE_ADDR)addr;
+ ret->kp.pre_handler = gtp_kp_pre_handler;
+ INIT_LIST_HEAD(&ret->action_list);
+
+ list_add(&ret->node, >p_list);
+
+out:
+ return ret;
+}
+
+static struct gtp_entry *gtp_list_find(ULONGEST num, ULONGEST addr)
+{
+ struct gtp_entry *tpe;
+ struct list_head *cur;
+
+ list_for_each(cur, >p_list) {
+ tpe = list_entry(cur, struct gtp_entry, node);
+ if (tpe->num == num && tpe->addr == addr)
+ return tpe;
+ }
+
+ return NULL;
+}
+
+static void gtp_list_release(void)
+{
+ struct gtp_entry *tpe;
+ struct list_head *cur, *next;
+
+ list_for_each_safe(cur, next, >p_list) {
+ struct action *ae;
+ struct list_head *cur_ae, *next_ae;
+
+ tpe = list_entry(cur, struct gtp_entry, node);
+
+ list_for_each_safe(cur_ae, next_ae, &tpe->action_list) {
+ ae = list_entry(cur_ae, struct action, node);
+ kfree(ae);
+ }
+
+ kfree(tpe);
+ }
+}
+
+static void gtp_frame_reset(void)
+{
+ gtp_frame_num = 0;
+ gtp_frame_r_start = gtp_frame;
+ gtp_frame_w_start = gtp_frame;
+ gtp_frame_w_end = gtp_frame + GTP_FRAME_SIZE;
+ gtp_frame_is_circular = 0;
+ gtp_frame_r_cache = NULL;
+ gtp_frame_current = NULL;
+}
+
+static int gtp_gdbrsp_qtinit(void)
+{
+ if (gtp_start)
+ return -EBUSY;
+
+ gtp_list_release();
+
+ if (gtp_frame)
+ gtp_frame_reset();
+
+ return 0;
+}
+
+static int gtp_gdbrsp_qtdp(char *pkg)
+{
+ int addnew = 1;
+ ULONGEST num, addr;
+ struct gtp_entry *tpe;
+
+ pr_devel("gtp_gdbrsp_qtdp: %s\n", pkg);
+
+ if (gtp_start)
+ return -EBUSY;
+
+ if (pkg[0] == '-') {
+ pkg++;
+ addnew = 0;
+ }
+
+ /* Get num and addr. */
+ if (pkg[0] == '\0')
+ return -EINVAL;
+ num = simple_strtoull(pkg, &pkg, 16);
+ if (pkg[0] == '\0')
+ return -EINVAL;
+ pkg++;
+ addr = simple_strtoull(pkg, &pkg, 16);
+ if (pkg[0] == '\0')
+ return -EINVAL;
+ pkg++;
+
+ tpe = gtp_list_find(num, addr);
+ if (addnew) {
+ if (tpe)
+ return -EINVAL;
+ if (pkg[0] == 'D')
+ return 0;
+ if (pkg[0] == '\0')
+ return -EINVAL;
+ pkg++;
+
+ tpe = gtp_list_add(num, addr);
+ if (tpe == NULL)
+ return -ENOMEM;
+
+ /* Get step and pass. */
+ if (pkg[0] == '\0')
+ return -EINVAL;
+ pkg++;
+ tpe->step = simple_strtoull(pkg, &pkg, 16);
+ if (pkg[0] == '\0')
+ return -EINVAL;
+ pkg++;
+ tpe->pass = simple_strtoull(pkg, &pkg, 16);
+ if (tpe->pass == 0)
+ tpe->nopass = 1;
+ } else if (tpe) {
+ /* Add action to tpe. */
+ int step_action = 0;
+
+ if (pkg[0] == 'S') {
+ pkg++;
+ step_action = 1;
+ /* XXX: Still not support step. */
+ return 1;
+ }
+ while (pkg[0]) {
+ struct action *ae;
+
+ switch (pkg[0]) {
+ case 'R':
+ /* XXX: reg_mask is ignore. */
+ ae = gtp_action_alloc(pkg[0]);
+ if (!ae)
+ return -ENOMEM;
+ pkg++;
+ ae->type = 'r';
+ ae->u.reg_mask = simple_strtoull(pkg, &pkg, 16);
+ break;
+ default:
+ /* XXX: Not support. */
+ return 1;
+ }
+
+ if (ae)
+ list_add(&ae->node, &tpe->action_list);
+ }
+ } else
+ return -EINVAL;
+
+ return 0;
+}
+
+static int gtp_gdbrsp_qtdisconnected(char *pkg)
+{
+ ULONGEST setting;
+
+ if (gtp_start)
+ return -EBUSY;
+ if (pkg[0] == '\0')
+ return -EINVAL;
+
+ setting = simple_strtoull(pkg, NULL, 16);
+ gtp_disconnected_tracing = (int) setting;
+
+ return 0;
+}
+
+static int gtp_gdbrsp_qtbuffer(char *pkg)
+{
+ if (strncmp("circular:", pkg, 9) == 0) {
+ ULONGEST setting;
+
+ pkg += 9;
+ if (pkg[0] == '\0')
+ return -EINVAL;
+ setting = simple_strtoull(pkg, NULL, 16);
+ gtp_circular = (int)setting;
+
+ return 0;
+ }
+
+ return 1;
+}
+
+static struct gtp_frame_head *gtp_frame_head_find(int num)
+{
+ struct gtp_frame_head *ret = NULL;
+ char *tmp;
+
+ if (gtp_frame_r_cache) {
+ tmp = gtp_frame_r_cache;
+
+ while (tmp < gtp_frame_w_start) {
+ switch (tmp[0]) {
+ case 'h':
+ ret = (struct gtp_frame_head *)(tmp + 1);
+ goto cache_check;
+ break;
+ case 'r':
+ tmp += GTP_FRAME_REG_SIZE;
+ break;
+ default:
+ goto cache_check;
+ break;
+ }
+ }
+
+cache_check:
+ if (ret && ret->frame_num == num)
+ goto out;
+ }
+
+ tmp = gtp_frame_r_start;
+ while (tmp < gtp_frame_w_start) {
+ switch (tmp[0]) {
+ case 'h':
+ ret = (struct gtp_frame_head *) (tmp + 1);
+ if (ret->frame_num == num)
+ goto out;
+ ret = NULL;
+ tmp += GTP_FRAME_HEAD_SIZE;
+ break;
+ case 'r':
+ tmp += GTP_FRAME_REG_SIZE;
+ break;
+ default:
+ goto out;
+ }
+ }
+
+out:
+ return ret;
+}
+
+static int gtp_gdbrsp_qtframe(char *pkg)
+{
+ if (gtp_start)
+ return -EBUSY;
+
+ if (strncmp(pkg, "pc:", 3) == 0) /* XXX */
+ return 1;
+ else if (strncmp(pkg, "tdp:", 4) == 0) /* XXX */
+ return 1;
+ else if (strncmp(pkg, "range:", 6) == 0) /* XXX */
+ return 1;
+ else if (strncmp(pkg, "outside:", 8) == 0) /* XXX */
+ return 1;
+ else {
+ ULONGEST num;
+ struct gtp_frame_head *ret;
+
+ if (pkg[0] == '\0')
+ return -EINVAL;
+ num = simple_strtoull(pkg, NULL, 16);
+
+ pr_devel("gtp_gdbrsp_qtframe: %d\n", (int) num);
+
+ if (((int) num) < 0) {
+ /* Return to current. */
+ gtp_frame_current = NULL;
+
+ return 0;
+ }
+ ret = gtp_frame_head_find((int) num);
+ if (ret) {
+ gtp_frame_current = ret;
+ gtp_frame_r_cache = (char *)ret;
+ gtp_frame_r_cache += sizeof(struct gtp_frame_head);
+ sprintf(gtp_rw_bufp, "F%xT%x",
+ gtp_frame_current->frame_num,
+ (unsigned int)gtp_frame_current->trace_num);
+ gtp_rw_size += strlen(gtp_rw_bufp);
+ gtp_rw_bufp += strlen(gtp_rw_bufp);
+ } else {
+ strcpy(gtp_rw_bufp, "F-1");
+ gtp_rw_bufp += 3;
+ gtp_rw_size += 3;
+ }
+ }
+
+ return 1;
+}
+
+static int gtp_gdbrsp_qtstop(void)
+{
+ struct gtp_entry *tpe;
+ struct list_head *cur;
+
+ if (!gtp_start)
+ return -EBUSY;
+
+ list_for_each(cur, >p_list) {
+ tpe = list_entry(cur, struct gtp_entry, node);
+ if (tpe->kpreg) {
+ unregister_kprobe(&tpe->kp);
+ tpe->kpreg = 0;
+ }
+ }
+
+ gtp_start = 0;
+
+ return 0;
+}
+
+static int gtp_gdbrsp_qtstart(void)
+{
+ struct gtp_entry *tpe;
+ struct list_head *cur;
+
+ if (gtp_start)
+ return -EBUSY;
+
+ if (!gtp_frame) {
+ gtp_frame = vmalloc(GTP_FRAME_SIZE);
+ if (!gtp_frame)
+ return -ENOMEM;
+
+ gtp_frame_reset();
+ }
+
+ list_for_each(cur, >p_list) {
+ tpe = list_entry(cur, struct gtp_entry, node);
+ if (!list_empty(&tpe->action_list)) {
+ int ret = register_kprobe(&tpe->kp);
+ if (ret < 0) {
+ gtp_gdbrsp_qtstop();
+ return ret;
+ }
+ tpe->kpreg = 1;
+ }
+ }
+
+ gtp_start = 1;
+
+ return 0;
+}
+
+static int gtp_gdbrsp_QT(char *pkg)
+{
+ int ret = 1;
+
+ pr_devel("gtp_gdbrsp_QT: %s\n", pkg);
+
+ if (strcmp("init", pkg) == 0)
+ ret = gtp_gdbrsp_qtinit();
+ else if (strncmp("DP:", pkg, 3) == 0)
+ ret = gtp_gdbrsp_qtdp(pkg + 3);
+ else if (strncmp("Disconnected:", pkg, 13) == 0)
+ ret = gtp_gdbrsp_qtdisconnected(pkg + 13);
+ else if (strncmp("Buffer:", pkg, 7) == 0)
+ ret = gtp_gdbrsp_qtbuffer(pkg + 7);
+ else if (strncmp("Frame:", pkg, 6) == 0)
+ ret = gtp_gdbrsp_qtframe(pkg + 6);
+ else if (strcmp("Start", pkg) == 0)
+ ret = gtp_gdbrsp_qtstart();
+ else if (strcmp("Stop", pkg) == 0)
+ ret = gtp_gdbrsp_qtstop();
+
+ return ret;
+}
+
+static unsigned char gtp_m_buffer[0xffff];
+
+static int gtp_gdbrsp_m(char *pkg)
+{
+ int i;
+ ULONGEST addr, len;
+
+ /* Get add and len. */
+ if (pkg[0] == '\0')
+ return -EINVAL;
+ addr = simple_strtoull(pkg, &pkg, 16);
+ if (pkg[0] != ',')
+ return -EINVAL;
+ pkg++;
+ len = simple_strtoull(pkg, &pkg, 16);
+ if (len == 0)
+ return -EINVAL;
+ len &= 0xffff;
+ len = (ULONGEST) min((int)(GTP_RW_BUFP_MAX / 2),
+ (int)len);
+
+ pr_devel("gtp_gdbrsp_m: addr = 0x%lx len = %d\n",
+ (unsigned long) addr, (int) len);
+
+ if (probe_kernel_read(gtp_m_buffer, (void *)(CORE_ADDR)addr,
+ (size_t)len))
+ return -EFAULT;
+
+ for (i = 0; i < (int)len; i++) {
+ sprintf(gtp_rw_bufp, "%02x", gtp_m_buffer[i]);
+ gtp_rw_bufp += 2;
+ gtp_rw_size += 2;
+ }
+
+ return 1;
+}
+
+static int gtp_gdbrsp_g(void)
+{
+ char *next;
+ struct gtp_frame_reg *fr = NULL;
+
+ if (GTP_RW_BUFP_MAX < GTP_REG_ASCII_SIZE)
+ return -E2BIG;
+
+ if (gtp_start || !gtp_frame_current)
+ goto empty_out;
+
+ /* Get the fr. */
+ next = gtp_frame_current->next;
+ if (next[0] == 'r') {
+ fr = (struct gtp_frame_reg *) (next + 1);
+ gtp_regs2ascii(&fr->regs, gtp_rw_bufp);
+ } else {
+empty_out:
+ memset(gtp_rw_bufp, '0', GTP_REG_ASCII_SIZE);
+ }
+ gtp_rw_bufp += GTP_REG_ASCII_SIZE;
+ gtp_rw_size += GTP_REG_ASCII_SIZE;
+ return 1;
+}
+
+static int gtp_open(struct inode *inode, struct file *file)
+{
+ int ret = 0;
+
+ down(>p_rw_lock);
+
+ if (gtp_rw_count > 0) {
+ ret = -EBUSY;
+ goto out;
+ }
+
+ gtp_read_ack = 0;
+ gtp_rw_buf = vmalloc(GTP_RW_MAX);
+ if (!gtp_rw_buf) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ gtp_rw_count++;
+
+out:
+ up(>p_rw_lock);
+ return ret;
+}
+
+static int gtp_release(struct inode *inode, struct file *file)
+{
+ down(>p_rw_lock);
+ vfree(gtp_rw_buf);
+
+ /* XXX: not handle gtp_disconnected_tracing. */
+ gtp_gdbrsp_qtstop();
+ gtp_gdbrsp_qtinit();
+ vfree(gtp_frame);
+ gtp_frame = NULL;
+
+ gtp_rw_count--;
+
+ up(>p_rw_lock);
+
+ return 0;
+}
+
+static long gtp_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+ /* This function will make GDB happy. */
+ pr_devel("gtp_ioctl: %x\n", cmd);
+
+ return 0;
+}
+
+static ssize_t gtp_write(struct file *file, const char __user *buf,
+ size_t size, loff_t *ppos)
+{
+ char *rsppkg = NULL;
+ int i, ret;
+ unsigned char csum = 0;
+
+ down(>p_rw_lock);
+
+ size = min_t(size_t, size, GTP_RW_MAX);
+ if (copy_from_user(gtp_rw_buf, buf, size)) {
+ size = -EFAULT;
+ goto out;
+ }
+
+ if (gtp_rw_buf[0] == '+' || gtp_rw_buf[0] == '-'
+ || gtp_rw_buf[0] == '\3')
+ goto out;
+
+ /* Check format and crc and get the rsppkg. */
+ for (i = 0; i < size - 2; i++) {
+ if (rsppkg == NULL) {
+ if (gtp_rw_buf[i] == '$')
+ rsppkg = gtp_rw_buf + i + 1;
+ } else {
+ if (gtp_rw_buf[i] == '#')
+ break;
+ else
+ csum += gtp_rw_buf[i];
+ }
+ }
+ if (rsppkg && gtp_rw_buf[i] == '#') {
+ /* Format is OK. Check crc. */
+ unsigned char c1, c2;
+
+ gtp_rw_buf[i] = '\0';
+
+ c1 = gtp_rw_buf[i+1];
+ c2 = gtp_rw_buf[i+2];
+ if (csum == (c1 << 4) + c2) {
+ pr_devel("gtp_write: crc error\n");
+ gtp_read_ack = '-';
+ goto out;
+ }
+ } else {
+ pr_devel("gtp_write: format error\n");
+ gtp_read_ack = '-';
+ goto out;
+ }
+ gtp_read_ack = '+';
+
+ pr_devel("gtp_write: %s\n", rsppkg);
+
+ /* Handle rsppkg and put return to gtp_rw_buf. */
+ gtp_rw_buf[0] = '$';
+ gtp_rw_bufp = gtp_rw_buf + 1;
+ gtp_rw_size = 0;
+ ret = 1;
+ switch (rsppkg[0]) {
+ case '?':
+ strcpy(gtp_rw_bufp, "S05");
+ gtp_rw_bufp += 3;
+ gtp_rw_size += 3;
+ break;
+ case 'g':
+ ret = gtp_gdbrsp_g();
+ break;
+ case 'm':
+ ret = gtp_gdbrsp_m(rsppkg + 1);
+ break;
+ case 'Q':
+ if (rsppkg[1] == 'T')
+ ret = gtp_gdbrsp_QT(rsppkg + 2);
+ break;
+ }
+ if (ret == 0) {
+ strcpy(gtp_rw_bufp, "OK");
+ gtp_rw_bufp += 2;
+ gtp_rw_size += 2;
+ } else if (ret < 0) {
+ sprintf(gtp_rw_bufp, "E%02x", -ret);
+ gtp_rw_bufp += 3;
+ gtp_rw_size += 3;
+ }
+
+ gtp_rw_bufp[0] = '#';
+ csum = 0;
+ for (i = 1; i < gtp_rw_size + 1; i++)
+ csum += gtp_rw_buf[i];
+ gtp_rw_bufp[1] = TOHEX(csum >> 4);
+ gtp_rw_bufp[2] = TOHEX(csum & 0x0f);
+ gtp_rw_bufp = gtp_rw_buf;
+ gtp_rw_size += 4;
+
+out:
+ up(>p_rw_lock);
+ return size;
+}
+
+static ssize_t gtp_read(struct file *file, char __user *buf, size_t size,
+ loff_t *ppos)
+{
+ if (size == 0)
+ return 0;
+
+ down(>p_rw_lock);
+
+ if (gtp_read_ack) {
+ int err = put_user(gtp_read_ack, buf);
+ if (err) {
+ size = -err;
+ goto out;
+ }
+ gtp_read_ack = 0;
+ size = 1;
+ goto out;
+ }
+
+ size = min(gtp_rw_size, size);
+
+ if (copy_to_user(buf, gtp_rw_bufp, size)) {
+ size = -EFAULT;
+ goto out;
+ }
+ gtp_rw_bufp += size;
+ gtp_rw_size -= size;
+
+out:
+ up(>p_rw_lock);
+ return size;
+}
+
+static const struct file_operations gtp_operations = {
+ .owner = THIS_MODULE,
+ .open = gtp_open,
+ .release = gtp_release,
+ .unlocked_ioctl = gtp_ioctl,
+ .compat_ioctl = gtp_ioctl,
+ .read = gtp_read,
+ .write = gtp_write,
+};
+
+struct dentry *gtp_dir;
+
+static int __init gtp_init(void)
+{
+ gtp_dir = debugfs_create_file("gtp", S_IFIFO | S_IRUSR | S_IWUSR, NULL,
+ NULL, >p_operations);
+ if (gtp_dir == NULL || gtp_dir == ERR_PTR(-ENODEV))
+ return -ENOMEM;
+
+ return 0;
+}
+
+static void __exit gtp_exit(void)
+{
+ if (gtp_dir)
+ debugfs_remove_recursive(gtp_dir);
+}
+
+module_init(gtp_init)
+module_exit(gtp_exit)
+
+MODULE_AUTHOR("Hui Zhu <teawater@gmail.com>");
+MODULE_LICENSE("GPL");
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1307,6 +1307,18 @@ config ASYNC_RAID6_TEST
If unsure, say N.
+config GTP
+ tristate "GDB tracepoint support"
+ depends on X86 || ARM || MIPS
+ select KPROBES
+ select DEBUG_FS
+ ---help---
+ KGTP is a realtime and lightweight Linux debugger and tracer.
+ It makes Linux Kernel supply a GDB remote debug interface.
+ Then GDB in current machine can debug and trace Linux through
+ GDB tracepoint and some other functions without stopping the
+ Linux Kernel.
+
source "samples/Kconfig"
source "lib/Kconfig.kgdb"
next prev parent reply other threads:[~2012-05-24 13:36 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-05-09 7:14 [PATCH]KGTP (Linux Kernel debugger and tracer) lite patch for review Hui Zhu
2012-05-09 14:05 ` Andi Kleen
2012-05-10 12:15 ` Hui Zhu
2012-05-10 17:38 ` Andi Kleen
2012-05-24 13:35 ` Hui Zhu [this message]
2012-05-24 15:07 ` Andi Kleen
2013-11-21 5:05 ` Hui Zhu
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=4FBE3938.30000@gmail.com \
--to=teawater@gmail.com \
--cc=andi@firstfloor.org \
--cc=geoff@infradead.org \
--cc=hch@infradead.org \
--cc=jason.wessel@windriver.com \
--cc=linux-kernel@vger.kernel.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.