UbiquitousOS
/
OpenCloudOS-Kernel
mirror of https://gitee.com/OpenCloudOS/OpenCloudOS-Kernel.git
11
0
Fork 0
Code Issues Projects Releases Wiki Activity

fs/ioctl: nvidia-smi process and memory display in docker 

Process information can not be displayed while running nvidia-smi
command in container.
The reason is nvidia-smi command using the host pid to retrieve
process information but the host pid does not exist in container
the nvidia-smi get nothing using the host pid.
sine we trap the nvidia-smi ioctl command to replace the host pid
with the guest, the nv_cmd_id and pad is changed, so update it by
nvidia module's version.

Fixes: fc35f2a69b37 ("ioctl: trap nvidia-smi command")
Signed-off-by: Huang Cun <cunhuang@tencent.com>
Reviewed-by: Jianping Liu <frankjpliu@tencent.com>
Reviewed-by: Yongliang Gao <leonylgao@tencent.com>
Signed-off-by: Jianping Liu <frankjpliu@tencent.com>
This commit is contained in:
Huang Cun 2024-04-08 11:03:33 +08:00 committed by Jianping Liu
parent c77c209977
commit dba2ad55b1
1 changed files with 60 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

69
fs/ioctl.c
View File

@ -705,6 +705,8 @@ int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
#define NV_CMD "nvidia-smi"
unsigned int nv_ioctl_id = 0xc020462a;
unsigned int nv_cmd_id_get_gram = 0x800203;
#define V525_105_17_PAD_NUM_17 17
#define V440_33_01_PAD_NUM_9 9
struct nv_get_pid_count {
u32 count;
@ -712,7 +714,7 @@ struct nv_get_pid_count {
struct pids {
u32 pid;
u32 pad[9];
u32 pad[0];
};
struct nv_get_gram {
@ -730,6 +732,47 @@ struct nv_cmd {
unsigned int sysctl_nvidia_smi_trap = 0;
static int get_pad_num_set_cmd_by_version(const char *version)
{
int pad_num = 0;
/* version 440.33.01 pad num is 9? */
if (strncmp(version, "525.", 4) >= 0) {
/* 525.105.17 have been cheched */
nv_cmd_id_get_gram = 0x2080018E;
pad_num = V525_105_17_PAD_NUM_17;
} else {
/* original supported version: 440.33.01 */
nv_cmd_id_get_gram = 0x800203;
pad_num = V440_33_01_PAD_NUM_9;
}
return pad_num;
}
static int read_nvidia_version(char *buf, int len)
{
struct file *file;
ssize_t bytes_read;
loff_t pos = 0;
file = filp_open("/sys/module/nvidia/version", O_RDONLY, 0);
if (IS_ERR(file)) {
pr_info("Cannot open /sys/module/nvidia/version\n");
return PTR_ERR(file);
}
bytes_read = kernel_read(file, buf, len - 1, &pos);
if (bytes_read < 0) {
pr_info("Cannot read nvidia version, ret=%zd\n", bytes_read);
filp_close(file, NULL);
return bytes_read;
}
buf[bytes_read] = '\0';
filp_close(file, NULL);
return 0;
}
static int convert_pid_to_container(int pid_in_host)
{
struct pid *ppid;
@ -793,15 +836,16 @@ static int get_nv_pid_count(struct nv_cmd *nv_cmd)
return count;
}
static void change_nv_pid(struct nv_cmd *nv_cmd, int count)
static void change_nv_pid(struct nv_cmd *nv_cmd, int count, int pad_num)
{
int len;
struct nv_get_gram *get_gram;
int n, i;
u32 guest_pid;
bool write;
int index;
len = sizeof(struct nv_get_gram) + sizeof(struct pids) * count;
len = sizeof(struct nv_get_gram) + (sizeof(u32) * (1 + pad_num))*count;
get_gram = (struct nv_get_gram *)kmalloc(len, GFP_KERNEL);
if (!get_gram) {
printk(KERN_ERR "change nv pid: malloc nv get gram failed\n");
@ -814,15 +858,16 @@ static void change_nv_pid(struct nv_cmd *nv_cmd, int count)
i = 0;
write = false;
while (i < get_gram->count) {
guest_pid = convert_pid_to_container(get_gram->pid[i].pid);
index = i * (pad_num + 1);
guest_pid = convert_pid_to_container(get_gram->pid[index].pid);
if (guest_pid > 0) {
/* If process run in other container then geust_pid will return 0. Then we should
* not change the pid
*/
printk(KERN_INFO "change nv pid: host pid:%d, container pid:%d\n", get_gram->pid[i].pid,
guest_pid);
pr_info("change nv pid: host pid:%d, container pid:%d\n",
get_gram->pid[index].pid, guest_pid);
write = true;
get_gram->pid[i].pid = guest_pid;
get_gram->pid[index].pid = guest_pid;
}
i++;
}
@ -843,7 +888,7 @@ int ksys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
int error;
struct fd f = fdget(fd);
struct nv_cmd nv_cmd;
int count;
int count, pad_num = 0;
if (!f.file)
return -EBADF;
@ -855,13 +900,19 @@ int ksys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
if ((sysctl_nvidia_smi_trap == 1) &&
(task_active_pid_ns(current)->level) && (nv_ioctl_id == cmd)) {
char buf[sizeof(current->comm)];
char version[20] = {0};
get_task_comm(buf, current);
if (strcmp(buf, NV_CMD) == 0 &&
read_nvidia_version(version, 20) == 0)
pad_num = get_pad_num_set_cmd_by_version(version);
if (pad_num == 0)
return error;
if ((strcmp(buf, NV_CMD) == 0) && (get_nv_cmd(arg, &nv_cmd) == 0) &&
(nv_cmd.cmd == nv_cmd_id_get_gram)) {
count = get_nv_pid_count(&nv_cmd);
if (count > 0)
change_nv_pid(&nv_cmd, count);
change_nv_pid(&nv_cmd, count, pad_num);
}
}