问题
I want to get information on the virtual memory layout of a process (stack, libc, heap, and code) for my current process in C (MacOS).
I read that in Linux you can do this by looking at /proc/<pid>/maps and that you can do this on the MacOS command line via the vmmap command line utility.
How can I do this in C? If possible I would like to get all pages in the virtual memory space pertaining to my process.
回答1:
You can use the mach_vm_region_info APIs to query this information from C. Here is an example from Jonathan Levin's book MacOS and iOS Internals:
#include <mach/vm_map.h>
#include <stdio.h>
#include <mach-o/dyld_images.h>
/**
* vmmap(1) clone for OS X and iOS
* -------------------------------
*
* This is a simple example of using the mach_vm_region_info APIs in order to
* obtain a process' (technically, a task's) virtual memory address space, in a
* manner akin to /proc/[pid]/maps on Linux.
*
* The process is simple - get the task port, then call mach_vm_region_info until
* you've exhausted the address space (in iOS this happens around 0x40000000,
* where the commpage is). On iOS 6, for some peculiar reason the task port is
* invalidated after each call, so the quick workaround here solves the problem
* by regetting the port. The actual mach error code to check for is in the header
* files, though the code simply tries regetting.
*
* N.B - For this code to work, you MUST provide the entitlements to allow
* task-for-pid to work, else you'll fail with error 5. The entitlements are in
* the output in Chapter 3, but for those of you who haven't bought the book, it would be:
*
--- Cut here
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
<key>com.apple.springboard.debugapplications</key>
<true/>
<key>get-task-allow</key>
<true/>
<key>proc_info-allow</key>
<true/>
<key>task_for_pid-allow</key>
<true/>
<key>run-unsigned-code</key>
<true/>
</dict>
</plist>
--- Ok, enough :-)
*
* so - copy the above XML to a file, say, "ent.xml", and be sure to run "ldid -Sent.xml vmmap"
* before trying to run this. You can download the binary (already thus signed) if you're lazy
* (and trust me, because you *will* need root on your i-Device for this)
*
* As the book clearly states, once you have the task port, the world is your oyster. You can
* control the entire virtual memory space, reading and writing it as you please. Stay tuned
* for the corrupt tool (which will be provided soon in binary form)
*
*/
int g_pid = 0; // required in iOS 6 (read below)
/* 03/08/13 - Added List of Mach-O images: */
struct dyld_image_info *g_dii = NULL;
int g_imageCount;
unsigned char *
readProcessMemory (int pid, mach_vm_address_t addr, mach_msg_type_number_t *size)
{
// Helper function to read process memory (a la Win32 API of same name)
// To make it easier for inclusion elsewhere, it takes a pid, and
// does the task_for_pid by itself. Given that iOS invalidates task ports
// after use, it's actually a good idea, since we'd need to reget anyway
task_t t;
task_for_pid(mach_task_self(),pid, &t);
mach_msg_type_number_t dataCnt = size;
vm_offset_t readMem;
// Use vm_read, rather than mach_vm_read, since the latter is different
// in iOS.
kern_return_t kr = vm_read(t, // vm_map_t target_task,
addr, // mach_vm_address_t address,
*size, // mach_vm_size_t size
&readMem, //vm_offset_t *data,
size); // mach_msg_type_number_t *dataCnt
if (kr) {
// DANG..
fprintf (stderr, "Unable to read target task's memory @%p - kr 0x%x\n" , addr, kr);
return NULL;
}
return ( (unsigned char *) readMem);
}
kern_return_t mach_vm_read (vm_map_t, mach_vm_address_t, mach_vm_size_t, vm_offset_t *, mach_msg_type_number_t *);
void
findListOfBinaries(task_t t, mach_vm_address_t addr, int size)
{
kern_return_t kr;
mach_msg_type_number_t dataCnt = size;
unsigned char *readData = readProcessMemory (g_pid, addr, &dataCnt);
int machsig = 0xfeedface;
// Checking only 0xfeedfa is a simple way to catch both 64-bit (facf) and 32-bit (face) headers
// Machine endianness is automatically taken care of, too..
if (readData && memcmp (readData + 1, ((unsigned char *) &machsig) + 1 , 3) == 0)
{
// This is a Mach header
int i = 0;
// A MUCH better way would be to iterate through the LC and find the name of dyld
// but this would require my machlib.c (closed source) and really get the same result.
// This works because on both iOS and OS X dyld is at /usr/lib.
for (i = 0; i <dataCnt; i++)
{
if (memcmp(readData+i, "lib/dyld", 8) == 0)
{
unsigned int dyld_all_image_infos_offset ;
int imageCount = 0;
memcpy (&dyld_all_image_infos_offset, readData+DYLD_ALL_IMAGE_INFOS_OFFSET_OFFSET, sizeof (unsigned int));
struct dyld_all_image_infos *dyldaii ;
// Safeguard: should check that dyld_all_image_infos_offset is < size..
if (dyld_all_image_infos_offset > size)
{
// This is to be expected, since the dyld_all_image_infos is in a data region
//printf ("Offset %x is greater than region size : %x\n", dyld_all_image_infos_offset, size);
dataCnt = sizeof(dyld_all_image_infos);
readData = readProcessMemory (g_pid, addr + dyld_all_image_infos_offset , &dataCnt);
if (!readData) { return;}
dyldaii = (struct dyld_all_image_infos *) readData;
}
else
{
dyldaii = (struct dyld_all_image_infos *) (readData +dyld_all_image_infos_offset);
}
printf ("Version: %d, %d images at offset %p\n",
dyldaii->version, dyldaii->infoArrayCount, dyldaii->infoArray);
// Go to dyldaii->infoArray address
imageCount = dyldaii->infoArrayCount;
dataCnt = imageCount * sizeof(struct dyld_image_info);
g_dii = (struct dyld_image_info *) malloc (dataCnt);
g_imageCount = imageCount;
readData = readProcessMemory(g_pid, dyldaii->infoArray, &dataCnt);
if (!readData) { return;}
struct dyld_image_info *dii = (struct dyld_image_info *) readData;
// We don't need i anymore, anyway
for (i = 0; i < imageCount; i++)
{
dataCnt = 1024;
char *imageName = readProcessMemory (g_pid, dii[i].imageFilePath, &dataCnt);
if (imageName) g_dii[i].imageFilePath = strdup(imageName);
else g_dii[i].imageFilePath = NULL;
g_dii[i].imageLoadAddress = dii[i].imageLoadAddress;
}
break;
}
}
}
}
/* End 03/08/13 */
char *
behavior_to_text (vm_behavior_t b)
{
switch (b)
{
case VM_BEHAVIOR_DEFAULT: return("default");
case VM_BEHAVIOR_RANDOM: return("random");
case VM_BEHAVIOR_SEQUENTIAL: return("fwd-seq");
case VM_BEHAVIOR_RSEQNTL: return("rev-seq");
case VM_BEHAVIOR_WILLNEED: return("will-need");
case VM_BEHAVIOR_DONTNEED: return("will-need");
case VM_BEHAVIOR_FREE: return("free-nowb");
case VM_BEHAVIOR_ZERO_WIRED_PAGES: return("zero-wire");
case VM_BEHAVIOR_REUSABLE: return("reusable");
case VM_BEHAVIOR_REUSE: return("reuse");
case VM_BEHAVIOR_CAN_REUSE: return("canreuse");
default: return ("?");
}
}
char *
protection_bits_to_rwx (vm_prot_t p)
{
// previous version of this somehow lost the "p&", always returning rwx..
static char returned[4];
returned[0] = (p &VM_PROT_READ ? 'r' : '-');
returned[1] = (p &VM_PROT_WRITE ? 'w' : '-');
returned[2] = (p & VM_PROT_EXECUTE ? 'x' : '-');
returned[3] = '\0';
// memory leak here. No biggy
return (strdup(returned));
}
const char *
unparse_inheritance (vm_inherit_t i)
{
switch (i)
{
case VM_INHERIT_SHARE:
return "share";
case VM_INHERIT_COPY:
return "copy";
case VM_INHERIT_NONE:
return "none";
default:
return "???";
}
}
macosx_debug_regions (task_t task, mach_vm_address_t address, int max)
{
kern_return_t kret;
mach_vm_address_t prev_address;
/* @TODO: warning - potential overflow here - gotta fix this.. */
vm_region_basic_info_data_t prev_info,info;
mach_vm_size_t size, prev_size;
mach_port_t object_name;
mach_msg_type_number_t count;
int nsubregions = 0;
int num_printed = 0;
count = VM_REGION_BASIC_INFO_COUNT_64;
kret = mach_vm_region (task, &address, &size, VM_REGION_BASIC_INFO,
(vm_region_info_t) &info, &count, &object_name);
if (kret != KERN_SUCCESS)
{
printf ("mach_vm_region: Error %d - %s", kret, mach_error_string(kret));
return;
}
memcpy (&prev_info, &info, sizeof (vm_region_basic_info_data_t));
prev_address = address;
prev_size = size;
nsubregions = 1;
for (;;)
{
int print = 0;
int done = 0;
address = prev_address + prev_size;
/* Check to see if address space has wrapped around. */
if (address == 0)
{
print = done = 1;
}
if (!done)
{
// Even on iOS, we use VM_REGION_BASIC_INFO_COUNT_64. This works.
count = VM_REGION_BASIC_INFO_COUNT_64;
kret =
mach_vm_region (task, &address, &size, VM_REGION_BASIC_INFO,
(vm_region_info_t) &info, &count, &object_name);
if (kret != KERN_SUCCESS)
{
/* iOS 6 workaround - attempt to reget the task port to avoiD */
/* "(ipc/send) invalid destination port" (1000003 or something) */
task_for_pid(mach_task_self(),g_pid, &task);
kret =
mach_vm_region (task, &address, &size, VM_REGION_BASIC_INFO,
(vm_region_info_t) &info, &count, &object_name);
}
if (kret != KERN_SUCCESS)
{
fprintf (stderr,"mach_vm_region failed for address %p - Error: %x\n", address,(kret));
size = 0;
if (address >= 0x4000000) return;
print = done = 1;
}
}
if (address != prev_address + prev_size)
print = 1;
if ((info.protection != prev_info.protection)
|| (info.max_protection != prev_info.max_protection)
|| (info.inheritance != prev_info.inheritance)
|| (info.shared != prev_info.reserved)
|| (info.reserved != prev_info.reserved))
print = 1;
if (print)
{
int print_size;
char *print_size_unit;
if (num_printed == 0)
printf ("Region ");
else
printf (" ... ");
findListOfBinaries(task, prev_address, prev_size);
/* Quick hack to show size of segment, which GDB does not */
print_size = prev_size;
if (print_size > 1024) { print_size /= 1024; print_size_unit = "K"; }
if (print_size > 1024) { print_size /= 1024; print_size_unit = "M"; }
if (print_size > 1024) { print_size /= 1024; print_size_unit = "G"; }
/* End Quick hack */
printf (" %p-%p [%d%s](%s/%s; %s, %s, %s) %s",
(prev_address),
(prev_address + prev_size),
print_size,
print_size_unit,
protection_bits_to_rwx (prev_info.protection),
protection_bits_to_rwx (prev_info.max_protection),
unparse_inheritance (prev_info.inheritance),
prev_info.shared ? "shared" : "private",
prev_info.reserved ? "reserved" : "not-reserved",
behavior_to_text (prev_info.behavior));
if (nsubregions > 1)
printf (" (%d sub-regions)", nsubregions);
printf ("\n");
prev_address = address;
prev_size = size;
memcpy (&prev_info, &info, sizeof (vm_region_basic_info_data_t));
nsubregions = 1;
num_printed++;
}
else
{
prev_size += size;
nsubregions++;
}
if ((max > 0) && (num_printed >= max))
{
printf ("Max %d num_printed %d\n", max, num_printed);
done = 1;
}
if (done)
break;
}
}
void
main(int argc, char **argv)
{
struct vm_region_basic_info vmr;
kern_return_t rc;
mach_port_t task;
mach_vm_size_t size = 8;
vm_region_info_t info = (vm_region_info_t) malloc(10000);
mach_msg_type_number_t info_count;
mach_port_t object_name;
mach_vm_address_t addr =1;
int pid;
if (!argv[1]) { printf ("Usage: %s <PID>\n"); exit (1);}
pid = atoi(argv[1]);
g_pid = pid; // req for iOS 6
rc = task_for_pid(mach_task_self(),pid, &task);
if (rc) { fprintf (stderr, "task_for_pid() failed with error %d - %s\n", rc, mach_error_string(rc)); exit(1); }
printf ("RC %d - Task: %d\n",rc, task);
macosx_debug_regions (task, addr, 1000);
int i ;
for ( i = 0; i < g_imageCount; i++)
{
printf("Image: %s loaded @%p\n",
g_dii[i].imageFilePath, g_dii[i].imageLoadAddress);
}
printf("Done\n");
}
I also found Julia Evans' blog post about recreating vmmap in Rust interesting and relevant to your question.
Another way you could do this is to call system with vmmap PID > file as the argument and then read the process mapping information from the file after system finishes.
来源：https://stackoverflow.com/questions/62090545/how-to-get-info-on-a-processs-virtual-memory-mappings-in-c-macos