* ssp.c (version): New global variable.

(longopts): Ditto.
(opts): Ditto.
(run_program): Correct cmd_line typos to cmdline.
(usage): New function.  Standardize usage output.  Add ssp.txt to --help
output.
(print_version): New function.
(main): Accommodate getopt.
This commit is contained in:
Christopher Faylor 2002-06-06 00:36:12 +00:00
parent ee964e2f97
commit c240a7fcde
2 changed files with 265 additions and 108 deletions

View File

@ -1,3 +1,14 @@
2002-06-03 Joshua Daniel Franklin <joshuadfranklin@yahoo.com>
* ssp.c (version): New global variable.
(longopts): Ditto.
(opts): Ditto.
(run_program): Correct cmd_line typos to cmdline.
(usage): New function. Standardize usage output. Add ssp.txt to
--help output.
(print_version): New function.
(main): Accommodate getopt.
2002-06-03 Joshua Daniel Franklin <joshuadfranklin@yahoo.com>
* umount.cc (version): New global variable.

View File

@ -20,53 +20,26 @@
#include <time.h>
#include <ctype.h>
#include <windows.h>
#include <getopt.h>
#ifdef __GNUC__
const char *help_text = "\
Usage: ssp [options] low_pc high_pc command...\n\
\n\
The SSP is a `single-step profiler' - it uses the debug API to\n\
single-step your program, noting *everything* your program runs, not\n\
just random places hit by random timer interrupts. You must specify\n\
the range of EIP values to profile. For example, you could profile\n\
just a function, or just a line of code, or the whole thing.\n\
Use \"objdump -h\" to find the start of .text and the section following\n\
it; this is what you want.\n\
\n\
There are many options to ssp. Since step-profiling makes your\n\
program run about 1,000 times slower than normal, it's best to\n\
understand all the options so that you can narrow down the parts\n\
of your program you need to single-step.\n\
\n\
-v = verbose messages about debug events.\n\
\n\
-d, -e = disable/enable single-stepping by default. Use\n\
OutputDebugString (\"ssp on\") to enable stepping, or \"ssp off\" to\n\
disable it. Thus, you can profile a single function call or block.\n\
\n\
-t = trace every EIP value to a file TRACE.SSP. This gets big *fast*.\n\
Use \"addr2line -C -f -s -e foo.exe < trace.ssp > lines.ssp\" and then\n\
\"perl cvttrace\" to convert to symbolic traces.\n\
\n\
-tc = trace every EIP value to the console. *Lots* slower.\n\
\n\
-s = trace sub-threads too. Dangerous if you have race conditions.\n\
\n\
-dll = enable dll profiling. A chart of relative DLL usage is\n\
produced after the run.\n\
\n\
Examples:\n\
ssp 0x401000 0x403000 hello.exe\n\
ssp -v -d -dll 0x401000 0x440000 foo.exe\n\
\n\
The output is a file \"gmon.out\" that can be read with gprof:\n\
gprof -b foo.exe\n\
\n\
See ssp.txt in the cygwin sources for more information.\n\
";
#else
char *help_text = "Usage: get cygwin!\n";
#endif
static const char version[] = "$Revision$";
static char *prog_name;
static struct option longopts[] =
{
{"console-trace", no_argument, NULL, 'c' },
{"disable", no_argument, NULL, 'd' },
{"enable", no_argument, NULL, 'e' },
{"help", no_argument, NULL, 'h' },
{"dll", no_argument, NULL, 'l' },
{"sub-threads", no_argument, NULL, 's' },
{"trace-eip", no_argument, NULL, 't' },
{"verbose", no_argument, NULL, 'v' },
{"version", no_argument, NULL, 'V' },
{NULL, 0, NULL, 0}
};
static char opts[] = "cdehlstvV";
#define KERNEL_ADDR 0x77000000
@ -102,7 +75,6 @@ typedef struct {
int low_pc=0, high_pc=0;
unsigned int last_pc=0, pc, last_sp=0, sp;
int total_cycles, count;
char *cmd_line;
HANDLE hProcess;
PROCESS_INFORMATION procinfo;
STARTUPINFO startup;
@ -330,7 +302,7 @@ run_program (char *cmdline)
memset (&startup, 0, sizeof (startup));
startup.cb = sizeof (startup);
if (!CreateProcess (0, cmd_line, 0, 0, 0,
if (!CreateProcess (0, cmdline, 0, 0, 0,
CREATE_NEW_PROCESS_GROUP
| CREATE_SUSPENDED
| DEBUG_PROCESS
@ -355,7 +327,7 @@ run_program (char *cmdline)
dll_info[0].base_address = 0;
dll_info[0].pcount = 0;
dll_info[0].scount = 0;
dll_info[0].name = cmd_line;
dll_info[0].name = cmdline;
num_dlls = 1;
SetThreadPriority (procinfo.hThread, THREAD_PRIORITY_IDLE);
@ -671,76 +643,250 @@ run_program (char *cmdline)
}
#define WHITE(s) while (isspace (*s)) s++
static void
usage (FILE * stream)
{
fprintf (stream , ""
"Usage: %s [options] low_pc high_pc command...\n"
" -c, --console-trace trace every EIP value to the console. *Lots* slower.\n"
" -d, --disable disable single-stepping by default; use\n"
" OutputDebugString (\"ssp on\") to enable stepping\n"
" -e, --enable enable single-stepping by default; use\n"
" OutputDebugString (\"ssp off\") to disable stepping\n"
" -h, --help output usage information and exit\n"
" -l, --dll enable dll profiling. A chart of relative DLL usage\n"
" is produced after the run.\n"
" -s, --sub-threads trace sub-threads too. Dangerous if you have\n"
" race conditions.\n"
" -t, --trace-eip trace every EIP value to a file TRACE.SSP. This\n"
" gets big *fast*.\n"
" -v, --verbose output verbose messages about debug events.\n"
" -V, --version output version information and exit\n"
"\n"
"Example: %s 0x401000 0x403000 hello.exe\n"
"\n"
"", prog_name, prog_name);
if (stream == stdout)
fprintf (stream , ""
"SSP - The Single Step Profiler\n"
"\n"
"Original Author: DJ Delorie <dj@redhat.com>\n"
"\n"
"The SSP is a program that uses the Win32 debug API to run a program\n"
"one ASM instruction at a time. It records the location of each\n"
"instruction used, how many times that instruction is used, and all\n"
"function calls. The results are saved in a format that is usable by\n"
"the profiling program \"gprof\", although gprof will claim the values\n"
"are seconds, they really are instruction counts. More on that later.\n"
"\n"
"Because the SSP was originally designed to profile the cygwin DLL, it\n"
"does not automatically select a block of code to report statistics on.\n"
"You must specify the range of memory addresses to keep track of\n"
"manually, but it's not hard to figure out what to specify. Use the\n"
"\"objdump\" program to determine the bounds of the target's \".text\"\n"
"section. Let's say we're profiling cygwin1.dll. Make sure you've\n"
"built it with debug symbols (else gprof won't run) and run objdump\n"
"like this:\n"
"\n"
" objdump -h cygwin1.dll\n"
"\n"
"It will print a report like this:\n"
"\n"
"cygwin1.dll: file format pei-i386\n"
"\n"
"Sections:\n"
"Idx Name Size VMA LMA File off Algn\n"
" 0 .text 0007ea00 61001000 61001000 00000400 2**2\n"
" CONTENTS, ALLOC, LOAD, READONLY, CODE, DATA\n"
" 1 .data 00008000 61080000 61080000 0007ee00 2**2\n"
" CONTENTS, ALLOC, LOAD, DATA\n"
" . . .\n"
"\n"
"The only information we're concerned with are the VMA of the .text\n"
"section and the VMA of the section after it (sections are usually\n"
"contiguous; you can also add the Size to the VMA to get the end\n"
"address). In this case, the VMA is 0x61001000 and the ending address\n"
"is either 0x61080000 (start of .data method) or 0x0x6107fa00 (VMA+Size\n"
"method).\n"
"\n"
"There are two basic ways to use SSP - either profiling a whole\n"
"program, or selectively profiling parts of the program.\n"
"\n"
"To profile a whole program, just run ssp without options. By default,\n"
"it will step the whole program. Here's a simple example, using the\n"
"numbers above:\n"
"\n"
" ssp 0x61001000 0x61080000 hello.exe\n"
"\n"
"This will step the whole program. It will take at least 8 minutes on\n"
"a PII/300 (yes, really). When it's done, it will create a file called\n"
"\"gmon.out\". You can turn this data file into a readable report with\n"
"gprof:\n"
"\n"
" gprof -b cygwin1.dll\n"
"\n"
"The \"-b\" means 'skip the help pages'. You can omit this until you're\n"
"familiar with the report layout. The gprof documentation explains\n"
"a lot about this report, but ssp changes a few things. For example,\n"
"the first part of the report reports the amount of time spent in each\n"
"function, like this:\n"
"\n"
"Each sample counts as 0.01 seconds.\n"
" %% cumulative self self total\n"
" time seconds seconds calls ms/call ms/call name\n"
" 10.02 231.22 72.43 46 1574.57 1574.57 strcspn\n"
" 7.95 288.70 57.48 130 442.15 442.15 strncasematch\n"
"\n"
"The \"seconds\" columns are really CPU opcodes, 1/100 second per opcode.\n"
"So, \"231.22\" above means 23,122 opcodes. The ms/call values are 10x\n"
"too big; 1574.57 means 157.457 opcodes per call. Similar adjustments\n"
"need to be made for the \"self\" and \"children\" columns in the second\n"
"part of the report.\n"
"\n"
"OK, so now we've got a huge report that took a long time to generate,\n"
"and we've identified a spot we want to work on optimizing. Let's say\n"
"it's the time() function. We can use SSP to selectively profile this\n"
"function by using OutputDebugString() to control SSP from within the\n"
"program. Here's a sample program:\n"
"\n"
" #include <windows.h>\n"
" main()\n"
" {\n"
" time_t t;\n"
" OutputDebugString(\"ssp on\");\n"
" time(&t);\n"
" OutputDebugString(\"ssp off\");\n"
" }\n"
"\n"
"Then, add the \"-d\" option to ssp to default to *disabling* profiling.\n"
"The program will run at full speed until the first OutputDebugString,\n"
"then step until the second.\n"
"\n"
" ssp -d 0x61001000 0x61080000 hello.exe\n"
"\n"
"You can then use gprof (as usual) to see the performance profile for\n"
"just that portion of the program's execution.\n"
"\n"
"There are many options to ssp. Since step-profiling makes your\n"
"program run about 1,000 times slower than normal, it's best to\n"
"understand all the options so that you can narrow down the parts\n"
"of your program you need to single-step.\n"
"\n"
"\"-v\" - verbose. This prints messages about threads starting and\n"
"stopping, OutputDebugString calls, DLLs loading, etc.\n"
"\n"
"\"-t\" and \"-c\" - tracing. With -t, *every* step's address is written\n"
"to the file \"trace.ssp\". This can be used to help debug functions,\n"
"since it can trace multiple threads. Clever use of scripts can match\n"
"addresses with disassembled opcodes if needed. Warning: creates\n"
"*huge* files, very quickly. \"-c\" prints each address to the console,\n"
"useful for debugging key chunks of assembler.\n"
"Use \"addr2line -C -f -s -e foo.exe < trace.ssp > lines.ssp\" and then\n"
"\"perl cvttrace\" to convert to symbolic traces.\n"
"\n"
"\"-s\" - subthreads. Usually, you only need to trace the main thread,\n"
"but sometimes you need to trace all threads, so this enables that.\n"
"It's also needed when you want to profile a function that only a\n"
"subthread calls. However, using OutputDebugString automatically\n"
"enables profiling on the thread that called it, not the main thread.\n"
"\n"
"\"-l\" - dll profiling. Generates a pretty table of how much time was\n"
"spent in each dll the program used. No sense optimizing a function in\n"
"your program if most of the time is spent in the DLL.\n"
"\n"
"I usually use the -v, -s, and -l options:\n"
"\n"
" ssp -v -s -l -d 0x61001000 0x61080000 hello.exe\n"
"\n");
if (stream == stderr)
fprintf (stream, "Try '%s --help' for more information.", prog_name);
exit (stream == stderr ? 1 : 0);
}
static void
print_version ()
{
const char *v = strchr (version, ':');
int len;
if (!v)
{
v = "?";
len = 1;
}
else
{
v += 2;
len = strchr (v, ' ') - v;
}
printf ("\
%s (cygwin) %.*s\n\
Single-Step Profiler\n\
Copyright 2000, 2001, 2002 Red Hat, Inc.\n\
Compiled on %s", prog_name, len, v, __DATE__);
}
int
main (int argc, char **argv)
{
int i, n;
int c, i;
int total_pcount, total_scount;
FILE *gmon;
setbuf (stdout, 0);
cmd_line = GetCommandLine ();
prog_name = strrchr (argv[0], '/');
if (prog_name == NULL)
prog_name = strrchr (argv[0], '\\');
if (prog_name == NULL)
prog_name = argv[0];
else
prog_name++;
/* strip off the program part */
n = sscanf (cmd_line, "%*s%n", &i);
cmd_line += i;
WHITE (cmd_line);
/* parse arguments. Ugly, but we need to preserve the tail of the line */
while (*cmd_line == '-')
while ((c = getopt_long (argc, argv, opts, longopts, NULL)) != EOF)
switch (c)
{
char opt[100];
n = sscanf (cmd_line, " %s%n", opt, &i);
cmd_line += i;
WHITE (cmd_line);
if (strcmp (opt, "-d") == 0)
{
printf ("stepping disabled; enable via OutputDebugString (\"ssp on\")\n");
stepping_enabled = 0;
}
if (strcmp (opt, "-e") == 0)
{
printf ("stepping enabled; disable via OutputDebugString (\"ssp off\")\n");
stepping_enabled = 1;
}
if (strcmp (opt, "-t") == 0)
{
printf ("tracing all $eip to trace.ssp\n");
tracing_enabled = 1;
}
if (strcmp (opt, "-tc") == 0)
{
printf ("tracing *all* $eip to the console\n");
trace_console = 1;
}
if (strcmp (opt, "-s") == 0)
{
printf ("tracing all sub-threads too, not just the main one\n");
trace_all_threads = 1;
}
if (strcmp (opt, "-dll") == 0)
{
printf ("profiling dll usage\n");
dll_counts = 1;
}
if (strcmp (opt, "-v") == 0)
{
printf ("verbose messages enabled\n");
verbose = 1;
}
case 'c':
printf ("tracing *all* $eip to the console\n");
trace_console = 1;
break;
case 'd':
printf ("stepping disabled; enable via OutputDebugString (\"ssp on\")\n");
stepping_enabled = 0;
break;
case 'e':
printf ("stepping enabled; disable via OutputDebugString (\"ssp off\")\n");
stepping_enabled = 1;
break;
case 'h':
usage (stdout);
break;
case 'l':
printf ("profiling dll usage\n");
dll_counts = 1;
break;
case 's':
printf ("tracing all sub-threads too, not just the main one\n");
trace_all_threads = 1;
break;
case 't':
printf ("tracing all $eip to trace.ssp\n");
tracing_enabled = 1;
break;
case 'v':
printf ("verbose messages enabled\n");
verbose = 1;
break;
case 'V':
print_version ();
exit (0);
default:
usage (stderr);
}
n = sscanf (cmd_line, " %i %i %n", &low_pc, &high_pc, &i);
if (n < 2)
{
fputs (help_text, stderr);
exit (1);
}
cmd_line += i;
if ( (argc - optind) < 3 )
usage (stderr);
sscanf (argv[optind++], "%i", &low_pc);
sscanf (argv[optind++], "%i", &high_pc);
if (low_pc > high_pc-8)
{
@ -752,9 +898,9 @@ main (int argc, char **argv)
memset (hits, 0, high_pc-low_pc+4);
fprintf (stderr, "prun: [%08x,%08x] Running `%s'\n",
low_pc, high_pc, cmd_line);
low_pc, high_pc, argv[optind]);
run_program (cmd_line);
run_program (argv[optind]);
hdr.lpc = low_pc;
hdr.hpc = high_pc;