DEFINE_SYSCALL(recvfrom, int, sockfd, void *, buf, size_t, len, int, flags, struct sockaddr *, src_addr, int *, addrlen)
{
log_info("recvfrom(%d, %p, %d, %x, %p, %p)", sockfd, buf, len, flags, src_addr, addrlen);
if (!mm_check_write(buf, len))
return -L_EFAULT;
if (src_addr)
{
if (!mm_check_write(addrlen, sizeof(*addrlen)))
return -L_EFAULT;
if (!mm_check_write(src_addr, *addrlen))
return -L_EFAULT;
}
struct file *f = vfs_get(sockfd);
if (!f)
return -L_EBADF;
int r;
if (!f->op_vtable->recvfrom)
{
log_error("recvfrom() not implemented.");
r = -L_ENOTSOCK;
}
else
r = f->op_vtable->recvfrom(f, buf, len, flags, src_addr, addrlen);
vfs_release(f);
return r;
}
DEFINE_SYSCALL(sched_getaffinity, pid_t, pid, size_t, cpusetsize, uint8_t *, mask)
{
log_info("sched_getaffinity(%d, %d, %p)\n", pid, cpusetsize, mask);
if (pid != 0)
{
log_error("pid != 0.\n");
return -ESRCH;
}
int bytes = (cpusetsize + 7) & ~7;
if (!mm_check_write(mask, bytes))
return -EFAULT;
for (int i = 0; i < bytes; i++)
mask[i] = 0;
/* TODO: Applications (i.e. ffmpeg) use this to detect the number of cpus and enable multithreading
* on cpu with multiple cores.
* Since we does not support multithreading at the time, we just report back one bit to let them
* think we only have one core and give up multithreading.
*/
mask[0] = 1;
#if 0
GROUP_AFFINITY affinity;
GetThreadGroupAffinity(GetCurrentThread(), &affinity);
int size = min(sizeof(uintptr_t), cpusetsize) * 8;
for (int i = 0; i < size; i++)
if (affinity.Mask & (1 << i))
mask[i / 8] |= 1 << i;
#endif
return sizeof(uintptr_t);
}
DEFINE_SYSCALL(clock_gettime, int, clk_id, struct timespec *, tp)
{
log_info("sys_clock_gettime(%d, 0x%p)", clk_id, tp);
if (!mm_check_write(tp, sizeof(struct timespec)))
return -L_EFAULT;
switch (clk_id)
{
case CLOCK_REALTIME:
{
FILETIME system_time;
win7compat_GetSystemTimePreciseAsFileTime(&system_time);
filetime_to_unix_timespec(&system_time, tp);
return 0;
}
case CLOCK_MONOTONIC:
case CLOCK_MONOTONIC_COARSE:
case CLOCK_MONOTONIC_RAW:
{
LARGE_INTEGER freq, counter;
QueryPerformanceFrequency(&freq);
QueryPerformanceCounter(&counter);
uint64_t ns = (double)counter.QuadPart / (double)freq.QuadPart * NANOSECONDS_PER_SECOND;
tp->tv_sec = ns / NANOSECONDS_PER_SECOND;
tp->tv_nsec = ns % NANOSECONDS_PER_SECOND;
return 0;
}
default:
return -L_EINVAL;
}
}
DEFINE_SYSCALL(getrlimit, int, resource, struct rlimit *, rlim)
{
log_info("getrlimit(%d, %p)\n", resource, rlim);
if (!mm_check_write(rlim, sizeof(struct rlimit)))
return -EFAULT;
switch (resource)
{
case RLIMIT_STACK:
rlim->rlim_cur = STACK_SIZE;
rlim->rlim_max = STACK_SIZE;
break;
case RLIMIT_NPROC:
log_info("RLIMIT_NPROC: return fake result.\n");
rlim->rlim_cur = 65536;
rlim->rlim_max = 65536;
break;
case RLIMIT_NOFILE:
rlim->rlim_cur = MAX_FD_COUNT;
rlim->rlim_max = MAX_FD_COUNT;
break;
default:
log_error("Unsupported resource: %d\n", resource);
return -EINVAL;
}
return 0;
}
DEFINE_SYSCALL(getrandom, void *, buf, size_t, buflen, unsigned int, flags)
{
log_info("getrandom(%p, %d, %x)", buf, buflen, flags);
if (!mm_check_write(buf, buflen))
return -L_EFAULT;
if (!RtlGenRandom(buf, buflen))
return 0;
return buflen;
}
static int mm_check_write_msghdr(struct msghdr *msg)
{
if (!mm_check_write(msg, sizeof(struct msghdr)))
return 0;
if (msg->msg_namelen && !mm_check_write(msg->msg_name, msg->msg_namelen))
return 0;
if (msg->msg_iovlen && !mm_check_write(msg->msg_iov, sizeof(struct iovec) * msg->msg_iovlen))
return 0;
if (msg->msg_controllen & !mm_check_write(msg->msg_control, msg->msg_controllen))
return 0;
for (int i = 0; i < msg->msg_iovlen; i++)
{
log_info("iov %d: [%p, %p)", i, msg->msg_iov[i].iov_base, (uintptr_t)msg->msg_iov[i].iov_base + msg->msg_iov[i].iov_len);
if (!mm_check_write(msg->msg_iov[i].iov_base, msg->msg_iov[i].iov_len))
return 0;
}
return 1;
}
DEFINE_SYSCALL(nanosleep, const struct timespec *, req, struct timespec *, rem)
{
log_info("nanosleep(0x%p, 0x%p)", req, rem);
if (!mm_check_read(req, sizeof(struct timespec)) || rem && !mm_check_write(rem, sizeof(struct timespec)))
return -L_EFAULT;
LARGE_INTEGER delay_interval;
delay_interval.QuadPart = 0ULL - (((uint64_t)req->tv_sec * 1000000000ULL + req->tv_nsec) / 100ULL);
NtDelayExecution(FALSE, &delay_interval);
return 0;
}
DEFINE_SYSCALL(accept4, int, sockfd, struct sockaddr *, addr, int *, addrlen, int, flags)
{
log_info("accept4(%d, %p, %p, %d)", sockfd, addr, addrlen, flags);
if (addr && !mm_check_write(addr, sizeof(struct sockaddr)))
return -L_EFAULT;
if (addrlen && !mm_check_write(addrlen, sizeof(int)))
return -L_EFAULT;
struct file *f = vfs_get(sockfd);
if (!f)
return -L_EBADF;
int r;
if (!f->op_vtable->accept4)
{
log_error("accept4() not implemented.");
r = -L_ENOTSOCK;
}
else
r = f->op_vtable->accept4(f, addr, addrlen, flags);
vfs_release(f);
return r;
}
DEFINE_SYSCALL(getpeername, int, sockfd, struct sockaddr *, addr, int *, addrlen)
{
log_info("getpeername(%d, %p, %p)", sockfd, addr, addrlen);
if (!mm_check_write(addrlen, sizeof(*addrlen)))
return -L_EFAULT;
if (!mm_check_write(addr, *addrlen))
return -L_EFAULT;
struct file *f = vfs_get(sockfd);
if (!f)
return -L_EBADF;
int r;
if (!f->op_vtable->getpeername)
{
log_error("getpeername() not implemented.");
r = -L_ENOTSOCK;
}
else
r = f->op_vtable->getpeername(f, addr, addrlen);
vfs_release(f);
return r;
}
DEFINE_SYSCALL(getsockopt, int, sockfd, int, level, int, optname, void *, optval, int *, optlen)
{
log_info("getsockopt(%d, %d, %d, %p, %p)", sockfd, level, optname, optval, optlen);
if (optlen && !mm_check_write(optlen, sizeof(*optlen)))
return -L_EFAULT;
if (optlen && !mm_check_write(optval, *optlen))
return -L_EFAULT;
struct file *f = vfs_get(sockfd);
if (!f)
return -L_EBADF;
int r;
if (!f->op_vtable->getsockopt)
{
log_error("getsockopt() not implemented.");
r = -L_ENOTSOCK;
}
else
r = f->op_vtable->getsockopt(f, level, optname, optval, optlen);
vfs_release(f);
return r;
}
DEFINE_SYSCALL(time, intptr_t *, c)
{
log_info("time(%p)", c);
if (c && !mm_check_write(c, sizeof(int)))
return -L_EFAULT;
FILETIME systime;
GetSystemTimeAsFileTime(&systime);
uint64_t t = filetime_to_unix_sec(&systime);
if (c)
*c = (intptr_t)t;
return t;
}
DEFINE_SYSCALL(getrusage, int, who, struct rusage *, usage)
{
log_info("getrusage(%d, %p)\n", who, usage);
if (!mm_check_write(usage, sizeof(struct rusage)))
return -EFAULT;
ZeroMemory(usage, sizeof(struct rusage));
switch (who)
{
default:
log_error("Unhandled who: %d.\n", who);
return -EINVAL;
}
}
DEFINE_SYSCALL(oldolduname, struct oldold_utsname *, buf)
{
if (!mm_check_write(buf, sizeof(struct oldold_utsname)))
return -EFAULT;
struct utsname newbuf;
sys_uname(&newbuf);
strncpy(buf->sysname, newbuf.sysname, __OLD_UTS_LEN + 1);
strncpy(buf->nodename, newbuf.nodename, __OLD_UTS_LEN + 1);
strncpy(buf->release, newbuf.release, __OLD_UTS_LEN + 1);
strncpy(buf->version, newbuf.version, __OLD_UTS_LEN + 1);
strncpy(buf->machine, newbuf.machine, __OLD_UTS_LEN + 1);
return 0;
}
DEFINE_SYSCALL(time, intptr_t *, c)
{
log_info("time(%p)\n", c);
if (c && !mm_check_write(c, sizeof(int)))
return -EFAULT;
SYSTEMTIME systime;
GetSystemTime(&systime);
uint64_t t = (uint64_t)systime.wSecond + (uint64_t)systime.wMinute * 60
+ (uint64_t)systime.wHour * 3600 + (uint64_t)systime.wDay * 86400
+ ((uint64_t)systime.wYear - 70) * 31536000 + (((uint64_t)systime.wYear - 69) / 4) * 86400
- (((uint64_t)systime.wYear - 1) / 100) * 86400 + (((uint64_t)systime.wYear + 299) / 400) * 86400;
if (c)
*c = (intptr_t)t;
return t;
}
DEFINE_SYSCALL(uname, struct utsname *, buf)
{
log_info("sys_uname(%p)\n", buf);
if (!mm_check_write(buf, sizeof(struct utsname)))
return -EFAULT;
/* Just mimic a reasonable Linux uname */
strcpy(buf->sysname, "Linux");
strcpy(buf->nodename, "ForeignLinux");
strcpy(buf->release, "3.15.0");
strcpy(buf->version, "3.15.0");
#ifdef _WIN64
strcpy(buf->machine, "x86_64");
#else
strcpy(buf->machine, "i686");
#endif
strcpy(buf->domainname, "GNU/Linux");
return 0;
}
DEFINE_SYSCALL(recv, int, sockfd, void *, buf, size_t, len, int, flags)
{
log_info("recv(%d, %p, %d, %x)", sockfd, buf, len, flags);
if (!mm_check_write(buf, len))
return -L_EFAULT;
struct file *f = vfs_get(sockfd);
if (!f)
return -L_EBADF;
int r;
if (!f->op_vtable->recvfrom)
{
log_error("recv() not implemented.");
r = -L_ENOTSOCK;
}
else
r = f->op_vtable->recvfrom(f, buf, len, flags, NULL, NULL);
vfs_release(f);
return r;
}
DEFINE_SYSCALL(clock_getres, int, clk_id, struct timespec *, res)
{
log_info("clock_getres(%d, 0x%p)", clk_id, res);
if (!mm_check_write(res, sizeof(struct timespec)))
return -L_EFAULT;
switch (clk_id)
{
case CLOCK_REALTIME:
{
ULONG coarse, fine, actual;
NtQueryTimerResolution(&coarse, &fine, &actual);
uint64_t ns = (uint64_t)actual * NANOSECONDS_PER_TICK;
res->tv_sec = ns / NANOSECONDS_PER_SECOND;
res->tv_nsec = ns % NANOSECONDS_PER_SECOND;
return 0;
}
case CLOCK_MONOTONIC:
case CLOCK_MONOTONIC_COARSE:
case CLOCK_MONOTONIC_RAW:
{
LARGE_INTEGER freq;
QueryPerformanceFrequency(&freq);
uint64_t ns = (double)1. / (double)freq.QuadPart;
if (ns == 0)
{
res->tv_sec = 0;
res->tv_nsec = 1;
}
else
{
res->tv_sec = ns / NANOSECONDS_PER_SECOND;
res->tv_nsec = ns % NANOSECONDS_PER_SECOND;
}
return 0;
}
default:
return -L_EINVAL;
}
}
DEFINE_SYSCALL(sysinfo, struct sysinfo *, info)
{
log_info("sysinfo(%p)\n", info);
if (!mm_check_write(info, sizeof(*info)))
return -EFAULT;
MEMORYSTATUSEX memory;
GlobalMemoryStatusEx(&memory);
info->uptime = (intptr_t)(GetTickCount64() / 1000ULL);
info->loads[0] = info->loads[1] = info->loads[2] = 0; /* TODO */
info->totalram = memory.ullTotalPhys / PAGE_SIZE;
info->freeram = memory.ullAvailPhys / PAGE_SIZE;
info->sharedram = 0;
info->bufferram = 0;
info->totalswap = memory.ullTotalPageFile / PAGE_SIZE;
info->freeswap = memory.ullAvailPageFile / PAGE_SIZE;
info->procs = 100; /* TODO */
info->totalhigh = 0;
info->freehigh = 0;
info->mem_unit = PAGE_SIZE;
RtlSecureZeroMemory(info->_f, sizeof(info->_f));
return 0;
}
static int dsp_ioctl(struct file *f, unsigned int cmd, unsigned long arg)
{
AcquireSRWLockExclusive(&f->rw_lock);
int r = 0;
struct dsp_file *dsp = (struct dsp_file *)f;
switch (cmd)
{
case SNDCTL_DSP_RESET:
{
log_info("SNDCTL_DSP_RESET.");
dsp_reset(dsp);
break;
}
case SNDCTL_DSP_SPEED:
{
if (!mm_check_read((int *)arg, sizeof(int)))
{
r = -L_EFAULT;
break;
}
int speed = *(int *)arg;
log_info("SNDCTL_DSP_SPEED: %d", speed);
DWORD old_speed = dsp->format.nSamplesPerSec;
dsp->format.nSamplesPerSec = speed;
if (!dsp_test_format(&dsp->format))
{
log_warning("Speed not supported.");
dsp->format.nSamplesPerSec = old_speed;
r = -L_EINVAL;
}
break;
}
case SNDCTL_DSP_STEREO:
{
if (!mm_check_read((int *)arg, sizeof(int)))
{
r = -L_EFAULT;
break;
}
int c = *(int *)arg;
log_info("SNDCTL_DSP_STEREO: %d", c);
if (c == 0)
dsp->format.nChannels = 1;
else if (c == 1)
dsp->format.nChannels = 2;
else
{
log_warning("Invalid argument (can only be 0 or 1).");
r = -L_EINVAL;
}
break;
}
case SNDCTL_DSP_SETFMT:
{
if (!mm_check_read((int *)arg, sizeof(int)))
{
r = -L_EFAULT;
break;
}
int fmt = *(int *)arg;
log_info("SNDCTL_DSP_SETFMT: 0x%x", fmt);
if (fmt == AFMT_S16_LE)
dsp->format.wBitsPerSample = 16;
else if (fmt == AFMT_U8)
dsp->format.wBitsPerSample = 8;
else
{
log_warning("Invalid argument (can only be AFMT_S16_LE or AFMT_U8).");
r = -L_EINVAL;
}
break;
}
case SNDCTL_DSP_GETFMTS:
{
if (!mm_check_write((int *)arg, sizeof(int)))
{
r = -L_EFAULT;
break;
}
log_info("SNDCTL_DSP_GETFMTS");
*(int *)arg = AFMT_U8 | AFMT_S16_LE;
break;
}
}
ReleaseSRWLockExclusive(&f->rw_lock);
return r;
}
static int load_elf(struct file *f, struct binfmt *binary)
{
struct elf_header *elf = binary->has_interpreter ? binary->interpreter : binary->executable;
/* Load ELF header */
f->op_vtable->pread(f, &elf->eh, sizeof(Elf_Ehdr), 0);
if (elf->eh.e_type != ET_EXEC && elf->eh.e_type != ET_DYN)
{
log_error("Only ET_EXEC and ET_DYN executables can be loaded.");
return -L_EACCES;
}
#ifdef _WIN64
if (elf->eh.e_machine != EM_X86_64)
{
log_error("Not an x86_64 executable.");
#else
if (elf->eh.e_machine != EM_386)
{
log_error("Not an i386 executable.");
#endif
return -L_EACCES;
}
/* Load program header table */
size_t phsize = (size_t)elf->eh.e_phentsize * (size_t)elf->eh.e_phnum;
char *pht = pht_storage;
f->op_vtable->pread(f, pht, phsize, elf->eh.e_phoff); /* TODO */
/* Find virtual address range */
elf->low = 0xFFFFFFFF;
elf->high = 0;
for (int i = 0; i < elf->eh.e_phnum; i++)
{
Elf_Phdr *ph = (Elf_Phdr *)&pht[elf->eh.e_phentsize * i];
if (ph->p_type == PT_LOAD)
{
elf->low = min(elf->low, ph->p_vaddr);
elf->high = max(elf->high, ph->p_vaddr + ph->p_memsz);
log_info("PT_LOAD: vaddr %p, size %p", ph->p_vaddr, ph->p_memsz);
}
else if (ph->p_type == PT_DYNAMIC)
log_info("PT_DYNAMIC: vaddr %p, size %p", ph->p_vaddr, ph->p_memsz);
else if (ph->p_type == PT_PHDR) /* Patch phdr pointer in PT_PHDR, glibc uses it to determine load offset */
ph->p_vaddr = (size_t)pht;
}
/* Find virtual address range for ET_DYN executable */
elf->load_base = 0;
if (elf->eh.e_type == ET_DYN)
{
size_t free_addr = mm_find_free_pages(elf->high - elf->low) * PAGE_SIZE;
if (!free_addr)
return -L_ENOMEM;
elf->load_base = free_addr - elf->low;
log_info("ET_DYN load offset: %p, real range [%p, %p)", elf->load_base, elf->load_base + elf->low, elf->load_base + elf->high);
}
#ifdef _WIN64
/* Unmap the pre-reserved executable region (see fork_init() for details) */
size_t region_start = 0x400000;
VirtualFree(region_start, 0, MEM_RELEASE); /* This will silently fail if it's not the intended case */
#endif
/* Map executable segments */
/* TODO: Directly use mmap() */
int load_base_set = 0;
for (int i = 0; i < elf->eh.e_phnum; i++)
{
Elf_Phdr *ph = (Elf_Phdr *)&pht[elf->eh.e_phentsize * i];
if (ph->p_type == PT_LOAD)
{
size_t addr = ph->p_vaddr & 0xFFFFF000;
size_t size = ph->p_memsz + (ph->p_vaddr & 0x00000FFF);
off_t offset_pages = ph->p_offset / PAGE_SIZE;
/* Note: In ET_DYN executables, all address are based upon elf->load_base.
* But in ET_EXEC executables, all address are absolute.
*/
int prot = 0;
if (ph->p_flags & PF_R)
prot |= PROT_READ;
if (ph->p_flags & PF_W)
prot |= PROT_WRITE;
if (ph->p_flags & PF_X)
prot |= PROT_EXEC;
if (elf->eh.e_type == ET_DYN)
addr += elf->load_base;
mm_mmap((void*)addr, size, PROT_READ | PROT_WRITE | PROT_EXEC,
MAP_ANONYMOUS | MAP_PRIVATE | MAP_FIXED | MAP_POPULATE, 0, NULL, 0);
char *vaddr = (char *)ph->p_vaddr;
if (elf->eh.e_type == ET_DYN)
vaddr += elf->load_base;
mm_check_write(vaddr, ph->p_filesz); /* Populate the memory, otherwise pread() will fail */
f->op_vtable->pread(f, vaddr, ph->p_filesz, ph->p_offset);
if (!binary->has_interpreter) /* This is not interpreter */
mm_update_brk((void*)(addr + size));
if (elf->eh.e_type == ET_EXEC && !load_base_set)
{
/* Record load base of first segment in ET_EXEC
* load_base will be used in run() to calculate various auxiliary vector pointers */
load_base_set = 1;
elf->load_base = addr;
}
}
}
/* Load interpreter if present */
for (int i = 0; i < elf->eh.e_phnum; i++)
{
Elf_Phdr *ph = (Elf_Phdr *)&pht[elf->eh.e_phentsize * i];
if (ph->p_type == PT_INTERP)
{
if (binary->has_interpreter) /* This is already an interpreter */
return -L_EACCES; /* Bad interpreter */
binary->has_interpreter = true;
char path[MAX_PATH];
f->op_vtable->pread(f, path, ph->p_filesz, ph->p_offset); /* TODO */
path[ph->p_filesz] = 0;
log_info("interpreter: %s", path);
struct file *fi;
int r = vfs_openat(AT_FDCWD, path, O_RDONLY, 0, 0, &fi);
if (r < 0)
return r;
if (!winfs_is_winfile(fi))
{
vfs_release(fi);
return -L_EACCES;
}
r = load_elf(fi, binary);
vfs_release(fi);
if (r < 0)
return -L_EACCES; /* Bad interpreter */
}
}
return 0;
}
#define MAX_SHEBANG_LINE 256
static int load_script(struct file *f, struct binfmt *binary)
{
/* Parse the shebang line */
int size = f->op_vtable->pread(f, binary->buffer_base, MAX_SHEBANG_LINE, 0);
char *p = binary->buffer_base, *end = p + size;
/* Skip shebang */
p += 2;
/* Skip spaces */
while (p < end && *p == ' ')
p++;
if (p == end)
return -L_EACCES;
const char *executable = p;
binary->argv0 = p;
while (p < end && *p != ' ' && *p != '\n')
p++;
if (p == end)
return -L_EACCES;
if (*p == '\n')
*p = 0; /* It has no argument */
else
{
*p++ = 0;
while (p < end && *p == ' ')
p++;
if (p == end)
return -L_EACCES;
if (*p != '\n')
{
/* It has an argument */
binary->argv1 = p;
while (p < end && *p != '\n')
p++;
if (p == end)
return -L_EACCES;
*p = 0;
}
}
binary->replace_argv0 = TRUE;
struct file *fe;
int r = vfs_openat(AT_FDCWD, executable, O_RDONLY, 0, 0, &fe);
if (r < 0)
return r;
if (!winfs_is_winfile(fe))
{
vfs_release(fe);
return -L_EACCES;
}
/* TODO: Recursive interpreters */
return load_elf(fe, binary);
}
