procfs: Support multiple reads of e.g. /proc/*/maps

Refs: #1021
Change-Id: If36e1a0f3f41f0215868daf578e96775d96a59a3
This commit is contained in:
Tomoki Shirasawa
2018-08-22 16:18:23 +09:00
committed by Masamichi Takagi
parent e531ee626e
commit 895a8c4099
10 changed files with 1042 additions and 230 deletions

View File

@ -67,6 +67,7 @@
#define SCD_MSG_PROCFS_DELETE 0x11
#define SCD_MSG_PROCFS_REQUEST 0x12
#define SCD_MSG_PROCFS_ANSWER 0x13
#define SCD_MSG_PROCFS_RELEASE 0x15
#define SCD_MSG_DEBUG_LOG 0x20

View File

@ -922,6 +922,303 @@ static const struct file_operations mckernel_forward = {
.write = mckernel_procfs_write,
};
#define PA_NULL (-1L)
struct mckernel_procfs_buffer_info {
unsigned long top_pa;
unsigned long cur_pa;
ihk_os_t os;
int pid;
char path[0];
};
struct mckernel_procfs_buffer {
unsigned long next_pa;
unsigned long pos;
unsigned long size;
char buf[0];
};
static int mckernel_procfs_buff_open(struct inode *inode, struct file *file)
{
struct mckernel_procfs_buffer_info *info;
int pid;
int ret;
char *path;
char *path_buf;
char *p;
ihk_os_t os;
#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 10, 0)
struct proc_dir_entry *dp = PDE(inode);
struct procfs_list_entry *e = dp->data;
#else
struct procfs_list_entry *e = PDE_DATA(inode);
#endif
os = osnum_to_os(e->osnum);
if (!os) {
return -EINVAL;
}
path_buf = kmalloc(PROCFS_NAME_MAX, GFP_KERNEL);
if (!path_buf) {
return -ENOMEM;
}
path = getpath(e, path_buf, PROCFS_NAME_MAX);
p = strchr(path, '/') + 1;
ret = sscanf(p, "%d/", &pid);
if (ret != 1) {
pid = -1;
}
info = kmalloc(sizeof(struct mckernel_procfs_buffer_info) +
strlen(path) + 1, GFP_KERNEL);
if (!info) {
kfree(path_buf);
return -ENOMEM;
}
info->top_pa = PA_NULL;
info->cur_pa = PA_NULL;
info->os = os;
info->pid = pid;
strcpy(info->path, path);
file->private_data = info;
kfree(path_buf);
return 0;
}
static int mckernel_procfs_buff_release(struct inode *inode, struct file *file)
{
struct mckernel_procfs_buffer_info *info = file->private_data;
int rc = 0;
if (!info) {
return -EIO;
}
file->private_data = NULL;
if (info->top_pa != PA_NULL) {
int ret;
struct procfs_read *r = NULL;
struct ikc_scd_packet isp;
int do_free;
r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
if (r == NULL) {
rc = -ENOMEM;
goto out;
}
memset(r, '\0', sizeof(struct procfs_read));
r->pbuf = info->top_pa;
r->ret = -EIO; /* default */
r->fname[0] = '\0';
isp.msg = SCD_MSG_PROCFS_RELEASE;
isp.ref = 0;
isp.arg = virt_to_phys(r);
isp.pid = 0;
rc = -EIO;
ret = mcctrl_ikc_send_wait(info->os, 0,
&isp, 5 * HZ, NULL, &do_free, 1, r);
if (!do_free && ret >= 0) {
ret = -EIO;
}
if (ret < 0) {
rc = ret;
if (ret == -ETIME) {
pr_info("%s: error: timeout (1 sec)\n",
__func__);
}
else if (ret == -ERESTARTSYS) {
rc = -ERESTART;
}
if (!do_free)
r = NULL;
goto out;
}
if (r->ret < 0) {
rc = r->ret;
goto out;
}
rc = 0;
out:
if (r)
kfree((void *)r);
}
kfree(info);
return rc;
}
static ssize_t mckernel_procfs_buff_read(struct file *file, char __user *ubuf,
size_t nbytes, loff_t *ppos)
{
struct mckernel_procfs_buffer_info *info = file->private_data;
unsigned long phys;
struct mckernel_procfs_buffer *buf;
int pos = *ppos;
ssize_t l = 0;
int done = 0;
ihk_os_t os;
if (nbytes <= 0 || *ppos < 0) {
return 0;
}
if (!info) {
return -EIO;
}
os = info->os;
if (info->top_pa == PA_NULL) {
int ret;
int pid = info->pid;
struct procfs_read *r = NULL;
struct ikc_scd_packet isp;
struct mcctrl_usrdata *udp = NULL;
struct mcctrl_per_proc_data *ppd = NULL;
int do_free;
udp = ihk_host_os_get_usrdata(os);
if (!udp) {
pr_err("%s: no MCCTRL data found for OS\n",
__func__);
return -EINVAL;
}
if (pid > 0) {
ppd = mcctrl_get_per_proc_data(udp, pid);
if (unlikely(!ppd)) {
pr_err("%s: no per-process structure for PID %d",
__func__, pid);
return -EINVAL;
}
}
r = kmalloc(sizeof(struct procfs_read), GFP_KERNEL);
if (r == NULL) {
l = -ENOMEM;
done = 1;
goto out;
}
memset(r, '\0', sizeof(struct procfs_read));
r->pbuf = PA_NULL;
r->ret = -EIO; /* default */
strncpy((char *)r->fname, info->path, PROCFS_NAME_MAX);
isp.msg = SCD_MSG_PROCFS_REQUEST;
isp.ref = 0;
isp.arg = virt_to_phys(r);
isp.pid = pid;
l = -EIO;
done = 1;
ret = mcctrl_ikc_send_wait(os,
(pid > 0) ? ppd->ikc_target_cpu : 0,
&isp, 5 * HZ, NULL, &do_free, 1, r);
if (!do_free && ret >= 0) {
ret = -EIO;
}
if (ret < 0) {
l = ret;
if (ret == -ETIME) {
pr_info("%s: error: timeout (1 sec)\n",
__func__);
}
else if (ret == -ERESTARTSYS) {
l = -ERESTART;
}
if (!do_free)
r = NULL;
goto out;
}
if (r->ret < 0) {
l = r->ret;
goto out;
}
done = 0;
l = 0;
info->top_pa = info->cur_pa = r->pbuf;
out:
if (ppd)
mcctrl_put_per_proc_data(ppd);
if (r)
kfree((void *)r);
}
if (info->cur_pa == PA_NULL) {
info->cur_pa = info->top_pa;
}
while (!done && info->cur_pa != PA_NULL) {
long bpos;
long bsize;
phys = ihk_device_map_memory(ihk_os_to_dev(os), info->cur_pa,
PAGE_SIZE);
#ifdef CONFIG_MIC
buf = ioremap_wc(phys, PAGE_SIZE);
#else
buf = ihk_device_map_virtual(ihk_os_to_dev(os), phys,
PAGE_SIZE, NULL, 0);
#endif
if (pos < buf->pos) {
info->cur_pa = info->top_pa;
goto rep;
}
if (pos >= buf->pos + buf->size) {
info->cur_pa = buf->next_pa;
goto rep;
}
bpos = pos - buf->pos;
bsize = (buf->pos + buf->size) - pos;
if (bsize > (nbytes - l)) {
bsize = nbytes - l;
}
if (copy_to_user(ubuf, buf->buf + bpos, bsize)) {
done = 1;
pos = *ppos;
l = -EFAULT;
}
else {
ubuf += bsize;
pos += bsize;
l += bsize;
if (l == nbytes) {
done = 1;
}
}
rep:
#ifdef CONFIG_MIC
iounmap(buf);
#else
ihk_device_unmap_virtual(ihk_os_to_dev(os), buf, PAGE_SIZE);
#endif
ihk_device_unmap_memory(ihk_os_to_dev(os), phys, PAGE_SIZE);
};
*ppos = pos;
return l;
}
static const struct file_operations mckernel_buff_io = {
.llseek = mckernel_procfs_lseek,
.read = mckernel_procfs_buff_read,
.write = NULL,
.open = mckernel_procfs_buff_open,
.release = mckernel_procfs_buff_release,
};
static const struct procfs_entry tid_entry_stuff[] = {
// PROC_REG("auxv", S_IRUSR, NULL),
// PROC_REG("clear_refs", S_IWUSR, NULL),
@ -931,10 +1228,10 @@ static const struct procfs_entry tid_entry_stuff[] = {
// PROC_LNK("exe", mckernel_readlink),
// PROC_REG("limits", S_IRUSR|S_IWUSR, NULL),
// PROC_REG("maps", S_IRUGO, NULL),
PROC_REG("mem", S_IRUSR|S_IWUSR, NULL),
PROC_REG("mem", 0600, NULL),
// PROC_REG("pagemap", S_IRUGO, NULL),
// PROC_REG("smaps", S_IRUGO, NULL),
PROC_REG("stat", S_IRUGO, NULL),
PROC_REG("stat", 0444, &mckernel_buff_io),
// PROC_REG("statm", S_IRUGO, NULL),
// PROC_REG("status", S_IRUGO, NULL),
// PROC_REG("syscall", S_IRUGO, NULL),
@ -943,26 +1240,26 @@ static const struct procfs_entry tid_entry_stuff[] = {
};
static const struct procfs_entry pid_entry_stuff[] = {
PROC_REG("auxv", S_IRUSR, NULL),
PROC_REG("auxv", 0400, &mckernel_buff_io),
/* Support the case where McKernel process retrieves its job-id under the Fujitsu TCS suite. */
// PROC_REG("cgroup", S_IXUSR, NULL),
// PROC_REG("clear_refs", S_IWUSR, NULL),
PROC_REG("cmdline", S_IRUGO, NULL),
PROC_REG("cmdline", 0444, &mckernel_buff_io),
// PROC_REG("comm", S_IRUGO|S_IWUSR, NULL),
// PROC_REG("coredump_filter", S_IRUGO|S_IWUSR, NULL),
PROC_REG("cpuset", S_IXUSR, NULL),
// PROC_REG("cpuset", S_IRUGO, NULL),
// PROC_REG("environ", S_IRUSR, NULL),
// PROC_LNK("exe", mckernel_readlink),
// PROC_REG("limits", S_IRUSR|S_IWUSR, NULL),
PROC_REG("maps", S_IRUGO, NULL),
PROC_REG("mem", S_IRUSR|S_IWUSR, NULL),
PROC_REG("pagemap", S_IRUGO, NULL),
PROC_REG("smaps", S_IRUGO, NULL),
// PROC_REG("stat", S_IRUGO, NULL),
PROC_REG("maps", 0444, &mckernel_buff_io),
PROC_REG("mem", 0600, NULL),
PROC_REG("pagemap", 0444, NULL),
// PROC_REG("smaps", S_IRUGO, NULL),
// PROC_REG("stat", 0444, &mckernel_buff_io),
// PROC_REG("statm", S_IRUGO, NULL),
PROC_REG("status", S_IRUGO, NULL),
PROC_REG("status", 0444, &mckernel_buff_io),
// PROC_REG("syscall", S_IRUGO, NULL),
PROC_DIR("task", S_IRUGO|S_IXUGO),
PROC_DIR("task", 0555),
// PROC_REG("wchan", S_IRUGO, NULL),
PROC_TERM
};
@ -970,14 +1267,14 @@ static const struct procfs_entry pid_entry_stuff[] = {
static const struct procfs_entry base_entry_stuff[] = {
// PROC_REG("cmdline", S_IRUGO, NULL),
#ifdef POSTK_DEBUG_ARCH_DEP_42 /* /proc/cpuinfo support added. */
PROC_REG("cpuinfo", S_IRUGO, NULL),
PROC_REG("cpuinfo", 0444, &mckernel_buff_io),
#else /* POSTK_DEBUG_ARCH_DEP_42 */
// PROC_REG("cpuinfo", S_IRUGO, NULL),
#endif /* POSTK_DEBUG_ARCH_DEP_42 */
// PROC_REG("meminfo", S_IRUGO, NULL),
// PROC_REG("pagetypeinfo",S_IRUGO, NULL),
// PROC_REG("softirq", S_IRUGO, NULL),
PROC_REG("stat", S_IRUGO, NULL),
PROC_REG("stat", 0444, &mckernel_buff_io),
// PROC_REG("uptime", S_IRUGO, NULL),
// PROC_REG("version", S_IRUGO, NULL),
// PROC_REG("vmallocinfo",S_IRUSR, NULL),

View File

@ -667,6 +667,7 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
break;
case SCD_MSG_PROCFS_REQUEST:
case SCD_MSG_PROCFS_RELEASE:
pckt.msg = SCD_MSG_PROCFS_ANSWER;
pckt.ref = packet->ref;
pckt.arg = packet->arg;

View File

@ -49,6 +49,7 @@
#define SCD_MSG_PROCFS_DELETE 0x11
#define SCD_MSG_PROCFS_REQUEST 0x12
#define SCD_MSG_PROCFS_ANSWER 0x13
#define SCD_MSG_PROCFS_RELEASE 0x15
#define SCD_MSG_DEBUG_LOG 0x20

View File

@ -40,6 +40,78 @@ extern int sprintf(char * buf, const char *fmt, ...);
extern int sscanf(const char * buf, const char * fmt, ...);
extern int scnprintf(char * buf, size_t size, const char *fmt, ...);
struct mckernel_procfs_buffer {
unsigned long next_pa;
unsigned long pos;
unsigned long size;
char buf[0];
};
#define PA_NULL (-1L)
static struct mckernel_procfs_buffer *buf_alloc(unsigned long *phys, long pos)
{
struct mckernel_procfs_buffer *buf;
buf = ihk_mc_alloc_pages(1, IHK_MC_AP_NOWAIT);
if (!buf)
return NULL;
buf->next_pa = PA_NULL;
buf->pos = pos;
buf->size = 0;
if (phys)
*phys = virt_to_phys(buf);
return buf;
}
static void buf_free(unsigned long phys)
{
struct mckernel_procfs_buffer *pbuf;
unsigned long next;
while (phys != PA_NULL) {
pbuf = phys_to_virt(phys);
next = pbuf->next_pa;
ihk_mc_free_pages(pbuf, 1);
phys = next;
}
}
static int buf_add(struct mckernel_procfs_buffer **top,
struct mckernel_procfs_buffer **cur, void *buf, int l)
{
int pos = 0;
int r;
int bufmax = PAGE_SIZE - sizeof(struct mckernel_procfs_buffer);
char *chr = buf;
if (!*top) {
*top = *cur = buf_alloc(NULL, 0);
if (!*top)
return -ENOMEM;
}
while (l) {
r = bufmax - (*cur)->size;
if (!r) {
*cur = buf_alloc(&(*cur)->next_pa, (*cur)->pos +
bufmax);
if (!*cur) {
buf_free(virt_to_phys(*top));
return -ENOMEM;
}
r = bufmax;
}
if (r > l) {
r = l;
}
memcpy((*cur)->buf + (*cur)->size, chr + pos, r);
l -= r;
pos += r;
(*cur)->size += r;
}
return 0;
}
static void
procfs_thread_ctl(struct thread *thread, int msg)
{
@ -85,13 +157,17 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
struct procfs_read *r;
int osnum = ihk_mc_get_osnum();
int rosnum, ret, pid, tid, ans = -EIO, eof = 0;
char *buf, *p;
char *buf, *p = NULL;
char *vbuf = NULL;
char *tmp = NULL;
struct mcs_rwlock_node_irqsave lock;
unsigned long offset;
int count;
int npages;
int readwrite = 0;
int err = -EIO;
struct mckernel_procfs_buffer *buf_top = NULL;
struct mckernel_procfs_buffer *buf_cur = NULL;
dprintf("process_procfs_request: invoked.\n");
@ -100,27 +176,55 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
dprintf("parg: %x\n", parg);
r = ihk_mc_map_virtual(parg, 1, PTATTR_WRITABLE | PTATTR_ACTIVE);
if (r == NULL) {
ihk_mc_unmap_memory(NULL, parg, sizeof(struct procfs_read));
kprintf("ERROR: process_procfs_request: got a null procfs_read structure.\n");
goto dataunavail;
goto err;
}
dprintf("r: %p\n", r);
dprintf("remote pbuf: %x\n", r->pbuf);
pbuf = ihk_mc_map_memory(NULL, r->pbuf, r->count);
dprintf("pbuf: %x\n", pbuf);
count = r->count + ((uintptr_t)pbuf & (PAGE_SIZE - 1));
npages = (count + (PAGE_SIZE - 1)) / PAGE_SIZE;
buf = ihk_mc_map_virtual(pbuf, npages, PTATTR_WRITABLE | PTATTR_ACTIVE);
dprintf("buf: %p\n", buf);
if (buf == NULL) {
kprintf("ERROR: process_procfs_request: got a null buffer.\n");
goto bufunavail;
if (rpacket->msg == SCD_MSG_PROCFS_RELEASE) {
struct mckernel_procfs_buffer *pbuf;
unsigned long phys;
unsigned long next;
for (phys = r->pbuf; phys != PA_NULL; phys = next) {
pbuf = phys_to_virt(phys);
next = pbuf->next_pa;
ihk_mc_free_pages(pbuf, 1);
}
r->ret = 0;
err = 0;
goto err;
}
readwrite = r->readwrite;
count = r->count;
if (r->pbuf == PA_NULL) {
tmp = ihk_mc_alloc_pages(1, IHK_MC_AP_NOWAIT);
if (!tmp)
goto err;
buf = tmp;
count = PAGE_SIZE;
}
else {
dprintf("remote pbuf: %x\n", r->pbuf);
pbuf = ihk_mc_map_memory(NULL, r->pbuf, r->count);
dprintf("pbuf: %x\n", pbuf);
count = r->count + ((uintptr_t)pbuf & (PAGE_SIZE - 1));
npages = (count + (PAGE_SIZE - 1)) / PAGE_SIZE;
vbuf = ihk_mc_map_virtual(pbuf, npages,
PTATTR_WRITABLE|PTATTR_ACTIVE);
dprintf("buf: %p\n", vbuf);
if (vbuf == NULL) {
ihk_mc_unmap_memory(NULL, pbuf, r->count);
kprintf("ERROR: %s: got a null buffer.\n", __func__);
goto err;
}
buf = vbuf;
readwrite = r->readwrite;
count = r->count;
dprintf("fname: %s, offset: %lx, count:%d.\n", r->fname,
r->offset, r->count);
}
offset = r->offset;
dprintf("fname: %s, offset: %lx, count:%d.\n", r->fname, r->offset, r->count);
/*
* check for "mcos%d/"
@ -200,35 +304,24 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
}
else if (!strcmp(p, "stat")) { /* "/proc/stat" */
extern int num_processors; /* kernel/ap.c */
char *p;
size_t remain;
int cpu;
if (offset > 0) {
ans = 0;
eof = 1;
goto end;
}
p = buf;
remain = count;
for (cpu = 0; cpu < num_processors; ++cpu) {
size_t n;
n = snprintf(p, remain, "cpu%d\n", cpu);
if (n >= remain) {
ans = -ENOSPC;
eof = 1;
goto end;
}
p += n;
snprintf(buf, count, "cpu%d\n", cpu);
if (buf_add(&buf_top, &buf_cur, buf, strlen(buf)) < 0)
goto err;
}
ans = p - buf;
eof = 1;
ans = 0;
goto end;
}
#ifdef POSTK_DEBUG_ARCH_DEP_42 /* /proc/cpuinfo support added. */
else if (!strcmp(p, "cpuinfo")) { /* "/proc/cpuinfo" */
ans = ihk_mc_show_cpuinfo(buf, count, offset, &eof);
ans = ihk_mc_show_cpuinfo(buf, count, 0, &eof);
if (ans < 0)
goto err;
if (buf_add(&buf_top, &buf_cur, buf, strlen(buf)) < 0)
goto err;
ans = 0;
goto end;
}
#endif /* POSTK_DEBUG_ARCH_DEP_42 */
@ -313,110 +406,51 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
*/
if (strcmp(p, "maps") == 0) {
struct vm_range *range;
#ifdef POSTK_DEBUG_TEMP_FIX_47 /* /proc/<pid>/maps 1024 byte over read fix. */
int left = PAGE_SIZE * 2;
#else /* POSTK_DEBUG_TEMP_FIX_47 */
int left = r->count - 1; /* extra 1 for terminating NULL */
#endif /* POSTK_DEBUG_TEMP_FIX_47 */
int written = 0;
char *_buf = buf;
#ifdef POSTK_DEBUG_TEMP_FIX_47 /* /proc/<pid>/maps 1024 byte over read fix. */
int len = 0;
char *tmp = NULL;
_buf = tmp = kmalloc(left, IHK_MC_AP_CRITICAL);
if (!tmp) {
kprintf("%s: error allocating /proc/self/maps buffer\n",
__FUNCTION__);
ans = 0;
goto end;
}
#endif /* POSTK_DEBUG_TEMP_FIX_47 */
#ifndef POSTK_DEBUG_TEMP_FIX_47 /* /proc/<pid>/maps 1024 byte over read fix. */
/* Starting from the middle of a proc file is not supported for maps */
if (offset > 0) {
ans = 0;
eof = 1;
goto end;
}
#endif /* POSTK_DEBUG_TEMP_FIX_47 */
ihk_mc_spinlock_lock_noirq(&vm->memory_range_lock);
range = lookup_process_memory_range(vm, 0, -1);
while (range) {
int written_now;
/* format is (from man proc):
* address perms offset dev inode pathname
* 08048000-08056000 r-xp 00000000 03:0c 64593 /usr/sbin/gpm
*/
written_now = snprintf(_buf, left,
"%012lx-%012lx %s%s%s%s %lx %lx:%lx %d\t\t\t%s\n",
range->start, range->end,
range->flag & VR_PROT_READ ? "r" : "-",
range->flag & VR_PROT_WRITE ? "w" : "-",
range->flag & VR_PROT_EXEC ? "x" : "-",
range->flag & VR_PRIVATE ? "p" : "s",
/* TODO: fill in file details! */
0UL,
0UL,
0UL,
0,
range->memobj && range->memobj->path ? range->memobj->path :
range->start ==
(unsigned long)vm->vdso_addr ? "[vdso]" :
range->start ==
(unsigned long)vm->vvar_addr ? "[vsyscall]" :
range->flag & VR_STACK ? "[stack]" :
range->start >= vm->region.brk_start &&
range->end <= vm->region.brk_end_allocated ?
"[heap]" :
snprintf(buf, count,
"%012lx-%012lx %s%s%s%s %lx %lx:%lx %d\t\t\t%s\n",
range->start, range->end,
range->flag & VR_PROT_READ ? "r" : "-",
range->flag & VR_PROT_WRITE ? "w" : "-",
range->flag & VR_PROT_EXEC ? "x" : "-",
range->flag & VR_PRIVATE ? "p" : "s",
/* TODO: fill in file details! */
0UL,
0UL,
0UL,
0,
range->memobj && range->memobj->path ?
range->memobj->path :
range->start == (unsigned long)vm->vdso_addr ?
"[vdso]" :
range->start == (unsigned long)vm->vvar_addr ?
"[vsyscall]" :
range->flag & VR_STACK ?
"[stack]" :
range->start >= vm->region.brk_start &&
range->end <= vm->region.brk_end_allocated ?
"[heap]" :
""
);
);
left -= written_now;
_buf += written_now;
written += written_now;
#ifdef POSTK_DEBUG_TEMP_FIX_47 /* /proc/<pid>/maps 1024 byte over read fix. */
if (left == 0) {
kprintf("%s(): WARNING: buffer too small to fill proc/maps\n",
__FUNCTION__);
break;
if (buf_add(&buf_top, &buf_cur, buf, strlen(buf)) < 0) {
ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
goto err;
}
#else /* POSTK_DEBUG_TEMP_FIX_47 */
if (left == 1) {
kprintf("%s(): WARNING: buffer too small to fill proc/maps\n",
__FUNCTION__);
break;
}
#endif /* POSTK_DEBUG_TEMP_FIX_47 */
range = next_process_memory_range(vm, range);
}
ihk_mc_spinlock_unlock_noirq(&vm->memory_range_lock);
#ifdef POSTK_DEBUG_TEMP_FIX_47 /* /proc/<pid>/maps 1024 byte over read fix. */
len = strlen(tmp);
if (r->offset < len) {
if (r->offset + r->count < len) {
ans = r->count;
} else {
eof = 1;
ans = len;
}
strncpy(buf, tmp + r->offset, ans);
} else if (r->offset == len) {
ans = 0;
eof = 1;
}
kfree(tmp);
#else /* POSTK_DEBUG_TEMP_FIX_47 */
ans = written;
eof = 1;
#endif /* POSTK_DEBUG_TEMP_FIX_47 */
ans = 0;
goto end;
}
@ -465,28 +499,16 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
extern int num_processors; /* kernel/ap.c */
struct vm_range *range;
unsigned long lockedsize = 0;
char *tmp;
char *bitmasks;
int bitmasks_offset = 0;
char *cpu_bitmask, *cpu_list, *numa_bitmask, *numa_list;
int len;
char *state;
tmp = kmalloc(8192, IHK_MC_AP_CRITICAL);
if (!tmp) {
kprintf("%s: error allocating /proc/self/status buffer\n",
__FUNCTION__);
ans = 0;
goto end;
}
bitmasks = kmalloc(BITMASKS_BUF_SIZE, IHK_MC_AP_CRITICAL);
if (!tmp) {
if (!bitmasks) {
kprintf("%s: error allocating /proc/self/status bitmaks buffer\n",
__FUNCTION__);
kfree(tmp);
ans = 0;
goto end;
goto err;
}
ihk_mc_spinlock_lock_noirq(&proc->vm->memory_range_lock);
@ -529,37 +551,43 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
state = "T (tracing stop)";
else if (proc->status == PS_EXITED)
state = "Z (zombie)";
sprintf(tmp,
"Pid:\t%d\n"
"Uid:\t%d\t%d\t%d\t%d\n"
"Gid:\t%d\t%d\t%d\t%d\n"
sprintf(buf,
"Pid:\t%d\n"
"Uid:\t%d\t%d\t%d\t%d\n"
"Gid:\t%d\t%d\t%d\t%d\n"
"State:\t%s\n"
"VmLck:\t%9lu kB\n"
"Cpus_allowed:\t%s\n"
"Cpus_allowed_list:\t%s\n"
"Mems_allowed:\t%s\n"
"Mems_allowed_list:\t%s\n",
proc->pid,
proc->ruid, proc->euid, proc->suid, proc->fsuid,
proc->rgid, proc->egid, proc->sgid, proc->fsgid,
"VmLck:\t%9lu kB\n",
proc->pid,
proc->ruid, proc->euid, proc->suid, proc->fsuid,
proc->rgid, proc->egid, proc->sgid, proc->fsgid,
state,
(lockedsize + 1023) >> 10,
cpu_bitmask, cpu_list, numa_bitmask, numa_list);
len = strlen(tmp);
if (r->offset < len) {
if (r->offset + r->count < len) {
ans = r->count;
} else {
eof = 1;
ans = len;
}
strncpy(buf, tmp + r->offset, ans);
} else if (r->offset == len) {
ans = 0;
eof = 1;
(lockedsize + 1023) >> 10);
if (buf_add(&buf_top, &buf_cur, buf, strlen(buf)) < 0) {
goto err;
}
sprintf(buf, "Cpus_allowed:\t%s\n", cpu_bitmask);
if (buf_add(&buf_top, &buf_cur, buf, strlen(buf)) < 0) {
kfree(bitmasks);
goto err;
}
sprintf(buf, "Cpus_allowed_list:\t%s\n", cpu_list);
if (buf_add(&buf_top, &buf_cur, buf, strlen(buf)) < 0) {
kfree(bitmasks);
goto err;
}
sprintf(buf, "Mems_allowed:\t%s\n", numa_bitmask);
if (buf_add(&buf_top, &buf_cur, buf, strlen(buf)) < 0) {
kfree(bitmasks);
goto err;
}
sprintf(buf, "Mems_allowed_list:\t%s\n", numa_list);
if (buf_add(&buf_top, &buf_cur, buf, strlen(buf)) < 0) {
kfree(bitmasks);
goto err;
}
kfree(tmp);
kfree(bitmasks);
ans = 0;
goto end;
}
@ -568,20 +596,10 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
*/
if (strcmp(p, "auxv") == 0) {
unsigned int limit = AUXV_LEN * sizeof(unsigned long);
unsigned int len = r->count;
if (r->offset < limit) {
if (limit < r->offset + r->count) {
len = limit - r->offset;
}
memcpy((void *)buf, ((char *) proc->saved_auxv) + r->offset, len);
ans = len;
if (r->offset + len == limit) {
eof = 1;
}
} else if (r->offset == limit) {
ans = 0;
eof = 1;
}
if (buf_add(&buf_top, &buf_cur, proc->saved_auxv, limit) < 0)
goto err;
ans = 0;
goto end;
}
@ -590,27 +608,17 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
*/
if (strcmp(p, "cmdline") == 0) {
unsigned int limit = proc->saved_cmdline_len;
unsigned int len = r->count;
if(!proc->saved_cmdline){
if (buf_add(&buf_top, &buf_cur, "", 0) < 0)
goto err;
ans = 0;
eof = 1;
goto end;
}
if (r->offset < limit) {
if (limit < r->offset + r->count) {
len = limit - r->offset;
}
memcpy((void *)buf, ((char *) proc->saved_cmdline) + r->offset, len);
ans = len;
if (r->offset + len == limit) {
eof = 1;
}
} else if (r->offset == limit) {
ans = 0;
eof = 1;
}
if (buf_add(&buf_top, &buf_cur, proc->saved_cmdline, limit) < 0)
goto err;
ans = 0;
goto end;
}
@ -622,8 +630,6 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
*/
if (!strcmp(p, "stat")) {
char tmp[1024];
int len;
/*
* pid (comm) state ppid
@ -638,7 +644,7 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
* cnswap exit_signal processor rt_priority
* policy delayacct_blkio_ticks guest_time cguest_time
*/
ans = sprintf(tmp,
ans = sprintf(buf,
"%d (%s) %c %d " // pid...
"%d %d %d %d " // pgrp...
"%u %lu %lu %lu " // flags...
@ -662,21 +668,10 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
0L, 0, thread->cpu_id, 0, // cnswap...
0, 0LL, 0L, 0L // policy...
);
dprintf("tmp=%s\n", tmp);
len = strlen(tmp);
if (r->offset < len) {
if (r->offset + r->count < len) {
ans = r->count;
} else {
eof = 1;
ans = len;
}
strncpy(buf, tmp + r->offset, ans);
} else if (r->offset == len) {
ans = 0;
eof = 1;
}
if (buf_add(&buf_top, &buf_cur, buf, strlen(buf)) < 0)
goto err;
ans = 0;
goto end;
}
@ -686,16 +681,24 @@ int process_procfs_request(struct ikc_scd_packet *rpacket)
kprintf("unsupported procfs entry: %d/%s\n", pid, p);
end:
ihk_mc_unmap_virtual(buf, npages);
dprintf("ret: %d, eof: %d\n", ans, eof);
r->ret = ans;
r->eof = eof;
err = 0;
bufunavail:
ihk_mc_unmap_memory(NULL, pbuf, r->count);
ihk_mc_unmap_virtual(r, 1);
dataunavail:
ihk_mc_unmap_memory(NULL, parg, sizeof(struct procfs_read));
if (r->pbuf == PA_NULL && buf_top)
r->pbuf = virt_to_phys(buf_top);
err:
if (vbuf) {
ihk_mc_unmap_virtual(vbuf, npages);
ihk_mc_unmap_memory(NULL, pbuf, r->count);
}
if (r) {
ihk_mc_unmap_virtual(r, 1);
ihk_mc_unmap_memory(NULL, parg, sizeof(struct procfs_read));
}
if (tmp) {
ihk_mc_free_pages(tmp, 1);
}
if(proc)
release_process(proc);

286
test/issues/1021/C1021.c Normal file
View File

@ -0,0 +1,286 @@
#include <stdlib.h>
#include <stdio.h>
#include <stdarg.h>
#include <string.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/mman.h>
#include <fcntl.h>
#include <errno.h>
#include <ctype.h>
int id;
int okcnt;
int ngcnt;
void *area;
void
ok(char *file, char *fmt, ...)
{
va_list ap;
printf("*** C1021T%02d %s ", id, file);
va_start(ap, fmt);
vprintf(fmt, ap);
va_end(ap);
printf("\n");
okcnt++;
}
void
ng(char *file, char *fmt, ...)
{
va_list ap;
printf("*** C1021T%02d %s ", id, file);
va_start(ap, fmt);
vprintf(fmt, ap);
va_end(ap);
printf("\n");
ngcnt++;
}
void
hex(char *bp, int len)
{
unsigned char *buf;
long l;
long p;
long zl = 0;
long zf = 1;
long zp = 0;
for (p = 0; p < len; p += 16) {
int i;
buf = (unsigned char *)bp + p;
l = 16;
if (p + 16 > len) {
l = len - p;
}
if (!zf) {
int zz = 0;
for (i = 0; i < l; i++) {
if (buf[i])
zz = 1;
}
if (l < 16 || !zz) {
zl += 16;
continue;
}
if (zl == 16) {
printf("%016lx 00000000 00000000 00000000 "
"00000000 *................*\n", zp);
}
else if (zl) {
printf(" %08lx - %08lx ZERO\n", zp, p);
}
}
zf = 0;
printf("%08lx ", p);
for (i = 0; i < 16; i++) {
if (i % 4 == 0)
printf(" ");
printf(i < l ? "%02x" : " ", buf[i]);
if (i < l && buf[i])
zf = 1;
}
printf(" *");
for (i = 0; i < 16; i++)
printf(i < l ? "%c" : " ",
isprint(buf[i]) ? buf[i] : '.');
printf("*\n");
zl = 0;
zp = p + 16;
}
}
void
sub(char *file, int mapsflag)
{
int fd;
int fd2;
char buf1[65536];
char buf2[65536];
char buf3[65536];
int n;
int rc;
int pos;
id++;
fd = open(file, O_RDONLY);
if (fd == -1) {
ng(file, "open %s", strerror(errno));
}
else {
ok(file, "open OK");
}
id++;
fd2 = dup(fd);
if (fd2 == -1) {
ng(file, "dup %s", strerror(errno));
}
else {
ok(file, "dup OK");
}
id++;
for (n = 0; (rc = read(fd, buf1 + n, 1)) == 1; n++);
if (rc == -1) {
ng(file, "read(1) %s", strerror(errno));
}
else if (mapsflag && n < 4096) {
ng(file, "read(1) short n=%d", n);
}
else {
ok(file, "read(1) OK n=%d", n);
}
id++;
rc = lseek(fd, 0L, SEEK_SET);
if (rc == -1) {
ng(file, "lseek %s", strerror(errno));
}
else {
ok(file, "lseek OK");
}
if (mapsflag)
munmap(area, 4096);
id++;
pos = 0;
while ((rc = read(fd, buf2 + pos, 1024)) > 0) {
pos += rc;
}
if (rc == -1) {
ng(file, "read(1) %s\n", strerror(errno));
}
else {
if (pos != n) {
ng(file, "read(1024) invalid size %d != %d", pos, n);
}
else if (memcmp(buf1, buf2, n) != 0) {
ng(file, "read(1024) invalid data");
hex(buf1, n);
hex(buf2, n);
}
else {
ok(file, "read(1024) OK");
}
}
id++;
rc = close(fd);
if (rc == -1) {
ng(file, "close %s", strerror(errno));
}
else {
ok(file, "close OK");
}
id++;
rc = read(fd2, buf3, n);
if (rc == -1) {
ng(file, "read(dup) EOF %s", strerror(errno));
}
else if (rc == 0) {
ok(file, "read(dup) EOF OK");
}
else {
ng(file, "read(dup) invalid position");
}
id++;
rc = lseek(fd2, 0L, SEEK_SET);
if (rc == -1) {
ng(file, "lseek(dup) %s", strerror(errno));
}
else {
ok(file, "lseek(dup) OK");
}
id++;
rc = read(fd2, buf3, n);
if (rc == -1) {
ng(file, "read(dup) %s", strerror(errno));
}
else if (rc != n) {
ng(file, "read(dup) too short");
}
else {
rc = read(fd2, buf3 + rc, n);
if (rc == -1) {
ng(file, "read(dup) %s", strerror(errno));
}
else if (rc != 0) {
ng(file, "read(dup) too long");
}
else if (memcmp(buf1, buf3, n) != 0) {
ng(file, "read(dup) invalid data");
hex(buf1, n);
hex(buf3, n);
}
else {
ok(file, "read(dup) OK");
}
}
id++;
rc = close(fd2);
if (rc == -1) {
ng(file, "close(dup) %s", strerror(errno));
}
else {
ok(file, "close(dup) OK");
}
}
int
main(int argc, char **argv)
{
int i;
int pid = getpid();
char file[1024];
for (i = 0; i < 512; i++) {
char *c;
if (i % 2) {
c = mmap(NULL, 4096, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
area = c;
}
else {
c = mmap(NULL, 4096, PROT_READ|PROT_WRITE,
MAP_SHARED|MAP_ANONYMOUS, -1, 0);
}
if (c == (void *)-1) {
printf("mmap error %d\n", errno);
exit(1);
}
*c = 1;
}
sub("/proc/stat", 0);
sprintf(file, "/proc/%d/auxv", pid);
sub(file, 0);
sprintf(file, "/proc/%d/cmdline", pid);
sub(file, 0);
sprintf(file, "/proc/%d/maps", pid);
sub(file, 1);
sprintf(file, "/proc/%d/status", pid);
sub(file, 0);
sprintf(file, "/proc/%d/task/%d/stat", pid, pid);
sub(file, 0);
if (ngcnt) {
printf("TEST FAILED OK=%d NG=%d\n", okcnt, ngcnt);
exit(1);
}
printf("TEST SUCCESS OK=%d\n", okcnt);
exit(0);
}

81
test/issues/1021/C1021.sh Executable file
View File

@ -0,0 +1,81 @@
#!/bin/sh
BIN=
SBIN=
LTP=
OSTEST=
BOOTPARAM="-c 1-7,17-23,9-15,25-31 -m 10G@0,10G@1"
if ! sudo ls /sys/kernel/debug | grep kmemleak > /dev/null 2>&1; then
echo kmemleak: not found >&2
exit 1
fi
if [ -f ../../../config.h ]; then
str=`grep "^#define BINDIR " ../../../config.h | head -1 | sed 's/^#define BINDIR /BINDIR=/'`
eval $str
fi
if [ "x$BINDIR" = x ];then
BINDIR="$BIN"
fi
if [ -f ../../../Makefile ]; then
str=`grep ^SBINDIR ../../../Makefile | head -1 | sed 's/ //g'`
eval $str
fi
if [ "x$SBINDIR" = x ];then
SBINDIR="$SBIN"
fi
if [ -f $HOME/ltp/testcases/bin/fork01 ]; then
LTPDIR=$HOME/ltp
fi
if [ "x$LTPDIR" = x ]; then
LTPDIR="$LTP"
fi
if [ "x$LTPDIR" != x ]; then
LTPDIR="$LTPDIR/testcases"
fi
if [ -f $HOME/ostest/bin/test_mck ]; then
OSTESTDIR=$HOME/ostest/
fi
if [ "x$OSTESTDIR" = x ]; then
OSTESTDIR="$OSTEST"
fi
if [ ! -x "$OSTESTDIR"/bin/test_mck ]; then
echo no ostest found $OSTEST >&2
exit 1
fi
TESTMCK="$OSTESTDIR/bin/test_mck"
if [ ! -x $SBINDIR/mcstop+release.sh ]; then
echo mcstop+release: not found >&2
exit 1
fi
echo -n "mcstop+release.sh ... "
sudo $SBINDIR/mcstop+release.sh
echo "done"
if [ ! -x $SBINDIR/mcreboot.sh ]; then
echo mcreboot: not found >&2
exit 1
fi
echo -n "mcreboot.sh $BOOTPARAM ... "
sudo $SBINDIR/mcreboot.sh $BOOTPARAM
echo "done"
if [ ! -x $BINDIR/mcexec ]; then
echo no mcexec found >&2
exit 1
fi
################################################################################
sudo sh -c 'echo clear > /sys/kernel/debug/kmemleak'
$BINDIR/mcexec ./C1021
sudo $SBINDIR/mcstop+release.sh
sudo sh -c 'echo scan > /sys/kernel/debug/kmemleak'
if sudo cat /sys/kernel/debug/kmemleak | tee C1021T71.kmemleak | grep 'mcctrl'; then
echo '*** C1021T61 NG (kmemleak)'
else
echo '*** C1021T61 OK (kmemleak)'
fi

View File

@ -0,0 +1,71 @@
Script started on Wed Aug 29 15:21:45 2018
bash-4.2$ make test
sh ./C1021.sh
mcstop+release.sh ... done
mcreboot.sh -c 1-7,17-23,9-15,25-31 -m 10G@0,10G@1 ... done
*** C1021T01 /proc/stat open OK
*** C1021T02 /proc/stat dup OK
*** C1021T03 /proc/stat read(1) OK n=158
*** C1021T04 /proc/stat lseek OK
*** C1021T05 /proc/stat read(1024) OK
*** C1021T06 /proc/stat close OK
*** C1021T07 /proc/stat read(dup) EOF OK
*** C1021T08 /proc/stat lseek(dup) OK
*** C1021T09 /proc/stat read(dup) OK
*** C1021T10 /proc/stat close(dup) OK
*** C1021T11 /proc/12455/auxv open OK
*** C1021T12 /proc/12455/auxv dup OK
*** C1021T13 /proc/12455/auxv read(1) OK n=144
*** C1021T14 /proc/12455/auxv lseek OK
*** C1021T15 /proc/12455/auxv read(1024) OK
*** C1021T16 /proc/12455/auxv close OK
*** C1021T17 /proc/12455/auxv read(dup) EOF OK
*** C1021T18 /proc/12455/auxv lseek(dup) OK
*** C1021T19 /proc/12455/auxv read(dup) OK
*** C1021T20 /proc/12455/auxv close(dup) OK
*** C1021T21 /proc/12455/cmdline open OK
*** C1021T22 /proc/12455/cmdline dup OK
*** C1021T23 /proc/12455/cmdline read(1) OK n=8
*** C1021T24 /proc/12455/cmdline lseek OK
*** C1021T25 /proc/12455/cmdline read(1024) OK
*** C1021T26 /proc/12455/cmdline close OK
*** C1021T27 /proc/12455/cmdline read(dup) EOF OK
*** C1021T28 /proc/12455/cmdline lseek(dup) OK
*** C1021T29 /proc/12455/cmdline read(dup) OK
*** C1021T30 /proc/12455/cmdline close(dup) OK
*** C1021T31 /proc/12455/maps open OK
*** C1021T32 /proc/12455/maps dup OK
*** C1021T33 /proc/12455/maps read(1) OK n=25401
*** C1021T34 /proc/12455/maps lseek OK
*** C1021T35 /proc/12455/maps read(1024) OK
*** C1021T36 /proc/12455/maps close OK
*** C1021T37 /proc/12455/maps read(dup) EOF OK
*** C1021T38 /proc/12455/maps lseek(dup) OK
*** C1021T39 /proc/12455/maps read(dup) OK
*** C1021T40 /proc/12455/maps close(dup) OK
*** C1021T41 /proc/12455/status open OK
*** C1021T42 /proc/12455/status dup OK
*** C1021T43 /proc/12455/status read(1) OK n=255
*** C1021T44 /proc/12455/status lseek OK
*** C1021T45 /proc/12455/status read(1024) OK
*** C1021T46 /proc/12455/status close OK
*** C1021T47 /proc/12455/status read(dup) EOF OK
*** C1021T48 /proc/12455/status lseek(dup) OK
*** C1021T49 /proc/12455/status read(dup) OK
*** C1021T50 /proc/12455/status close(dup) OK
*** C1021T51 /proc/12455/task/12455/stat open OK
*** C1021T52 /proc/12455/task/12455/stat dup OK
*** C1021T53 /proc/12455/task/12455/stat read(1) OK n=92
*** C1021T54 /proc/12455/task/12455/stat lseek OK
*** C1021T55 /proc/12455/task/12455/stat read(1024) OK
*** C1021T56 /proc/12455/task/12455/stat close OK
*** C1021T57 /proc/12455/task/12455/stat read(dup) EOF OK
*** C1021T58 /proc/12455/task/12455/stat lseek(dup) OK
*** C1021T59 /proc/12455/task/12455/stat read(dup) OK
*** C1021T60 /proc/12455/task/12455/stat close(dup) OK
TEST SUCCESS OK=60
*** C1021T61 OK (kmemleak)
bash-4.2$ exit
exit
Script done on Wed Aug 29 15:22:55 2018

13
test/issues/1021/Makefile Normal file
View File

@ -0,0 +1,13 @@
CC=gcc
TARGET=C1021
all:: $(TARGET)
C1021: C1021.c
$(CC) -o C1021 C1021.c -Wall -g
test:: $(TARGET)
sh ./C1021.sh
clean::
rm -f *.o $(TARGET)

58
test/issues/1021/README Normal file
View File

@ -0,0 +1,58 @@
【Issue#1021 動作確認】
□ テスト内容
1. procfs ファイルに対するファイルオペレーションのテスト
Issue#1021 の対応において、procfs の以下のファイル処理を変更している。
/proc/stat
/proc/pid/auxv
/proc/pid/cmdline
/proc/pid/maps
/proc/pid/status
/proc/pid/task/tid/stat
これらのファイルに対するファイルオペレーションとして、以下をテストする。
1) ファイルをopen(2)できること。
2) ファイルディスクリプタをdup(2)できること。
3) 1バイト単位にファイル終端までread(2)できること。
4) lseek(2) できること。(※2)
5) lseek(2) 後に、1024バイト単位に read(2) できること。最初の read(2)と内容が
一致していること。
6) close(2) できること。
7) dup(2) したファイルディスクリプタが EOF になっていること (5 の read(2) の
影響)。
8) dup(2) したファイルディスクリプタを lseek(2) できること。
9) dup(2) したファイルディスクリプタを read(2) し、ファイル全体を 1 回の read(2)
で読み込むことができること。また、最初の read(2) と内容が一致していること。
10) dup(2) したファイルを close(2) できること。
テストケースは以下の通りである。
C1021T01-C1021T10 /proc/stat に対する上記 1) - 10) のテスト
C1021T11-C1021T20 /proc/pid/auxv に対する上記 1) - 10) のテスト
C1021T21-C1021T30 /proc/pid/cmdline に対する上記 1) - 10) のテスト
C1021T31-C1021T40 /proc/pid/maps に対する上記 1) - 10) のテスト (※1, ※2)
C1021T41-C1021T50 /proc/pid/status に対する上記 1) - 10) のテスト
C1021T51-C1021T60 /proc/pid/rask/tid/stat に対する上記 1) - 10) のテスト
※1 /proc/pid/maps はユーザプログラムのメモリの使い方により非常に大きなファイル
になることがあるので、予め mmap を複数回行って4kB以上の read が発生するよう
にしておく。(McKernel 内のバッファが複数ページになる場合のテストを兼ねる)。
※2 /proc/pid/maps の読み込み中に情報が変化しても後続の read(2) に影響しない
ことを確認するため、/proc/pid/maps の lseek(2) 後に munmap(2) を行い、
McKernel の内部情報を変化させる。(後続の read(2) では、munmap 前の情報を
読み込む仕様)。
2. メモリリークが発生していないことの確認
C1021T61 kmemleak を用いて mcctrl の procfs 処理がメモリリークを起こして
いないことを確認する。
□ 実行手順
$ make test
実行できない場合は、C1021.shの以下の行を適切に書き換えた後に実行。
BIN= mcexec が存在するパス
SBIN= mcreboot.sh が存在するパス
LTP= LTP が存在するパス
□ 実行結果
C1021.txt 参照。
全ての項目が OK となっていることを確認。