packages/dpu-utilities
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

!36 Sync patches from source

Browse Source 
From: @swf504 
Reviewed-by: @minknov 
Signed-off-by: @minknov
This commit is contained in:
openeuler-ci-bot 2023-06-12 11:42:04 +00:00 committed by Gitee
commit 05d2a91198
No known key found for this signature in database
GPG Key ID: 173E9B9CA92EEF8F
15 changed files with 1983 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

547
0001-rewrite-client-rexec_run.patch Normal file
View File

@ -0,0 +1,547 @@
From 47fdab7bf180b058f6bbed10dd17e9a4c784eecc Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Thu, 1 Jun 2023 15:46:05 +0800
Subject: rewrite client rexec_run
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/rexec/rexec.c | 256 +++++++++++++++++++++++++++++---------
qtfs/rexec/rexec.h | 8 ++
qtfs/rexec/rexec_server.c | 51 +++++---
3 files changed, 236 insertions(+), 79 deletions(-)
diff --git a/qtfs/rexec/rexec.c b/qtfs/rexec/rexec.c
index 4dd206d..489ebec 100644
--- a/qtfs/rexec/rexec.c
+++ b/qtfs/rexec/rexec.c
@@ -42,6 +42,21 @@
#define REXEC_MSG_LEN 1024
FILE *rexec_logfile = NULL;
+struct rexec_global_var {
+ int rexec_hs_fd[2];
+};
+
+struct rexec_global_var g_rexec;
+
+
+struct rexec_client_event {
+ int fd;
+ int outfd; // for stdin out err and other pipe
+ int (*handler)(struct rexec_client_event *);
+ int *exit_status;
+ int *pidfd;
+};
+
#define REXEC_PIDMAP_PATH "/var/run/rexec/pids"
#define REXEC_PIDMAP_PATH_LEN 64
#define REXEC_PID_LEN 16
@@ -84,37 +99,39 @@ static int rexec_msg_fill_argv(int argc, char *argv[], char *msg)
return offset;
}
-static int rexec_io(int infd, int outfd, char *buf, int buflen)
+static int rexec_io(struct rexec_client_event *evt)
{
+#define MAX_MSG_LEN 256
+ char buf[MAX_MSG_LEN];
int len;
int ret;
- while ((len = read(infd, buf, buflen)) > 0) {
- ret = write(outfd, buf, len);
+ while ((len = read(evt->fd, buf, MAX_MSG_LEN)) > 0) {
+ ret = write(evt->outfd, buf, len);
if (ret <= 0) {
- rexec_err("Read from fd:%d len:%d write to fd:%d failed ret:%d", infd, len, outfd, ret);
- return -1;
+ rexec_err("Read from fd:%d len:%d write to fd:%d failed ret:%d", evt->fd, len, evt->outfd, ret);
+ return REXEC_EVENT_EXIT;
}
if (ret != len) {
- rexec_err("Read from fd:%d len:%d but write to fd:%d ret:%d", infd, len, outfd, ret);
+ rexec_err("Read from fd:%d len:%d but write to fd:%d ret:%d", evt->fd, len, evt->outfd, ret);
}
}
- return 0;
+ return REXEC_EVENT_OK;
}
// return -1 means process exit.
-static int rexec_conn_msg(int connfd, int *exit_status, int *pidfd)
+static int rexec_conn_msg(struct rexec_client_event *evt)
{
struct rexec_msg head;
- int ret = recv(connfd, &head, sizeof(struct rexec_msg), MSG_WAITALL);
+ int ret = recv(evt->fd, &head, sizeof(struct rexec_msg), MSG_WAITALL);
if (ret <= 0) {
rexec_err("Rexec conn recv err:%d errno:%d", ret, errno);
- return -1;
+ return REXEC_EVENT_EXIT;
}
switch (head.msgtype) {
case REXEC_KILL:
- *exit_status = head.exit_status;
+ *evt->exit_status = head.exit_status;
rexec_err("Rexec conn recv kill msg, exit:%d now.", head.exit_status);
- return -1;
+ return REXEC_EVENT_EXIT;
case REXEC_PIDMAP: {
int mypid = getpid();
int peerpid = head.pid;
@@ -122,9 +139,9 @@ static int rexec_conn_msg(int connfd, int *exit_status, int *pidfd)
char buf[REXEC_PID_LEN] = {0};
int fd;
int err;
- if (*pidfd > 0) {
+ if (*evt->pidfd > 0) {
rexec_err("Rexec pidmap msg > 1 error.");
- return 0;
+ return REXEC_EVENT_OK;
}
sprintf(path, "%s/%d", REXEC_PIDMAP_PATH, mypid);
fd = open(path, O_CREAT|O_WRONLY, 0600);
@@ -133,23 +150,41 @@ static int rexec_conn_msg(int connfd, int *exit_status, int *pidfd)
mypid, peerpid, path, fd);
break;
}
- *pidfd = fd;
+ *evt->pidfd = fd;
if ((err = flock(fd, LOCK_EX)) != 0) {
rexec_err("Rexec flock file:%s failed, errno:%d rexec exit.", path, err);
- return -1;
+ return REXEC_EVENT_EXIT;
}
if ((err = ftruncate(fd, 0)) != 0) {
rexec_err("Rexec pidmap file:%s clear failed errno:%d rexec exit.", path, err);
- return -1;
+ return REXEC_EVENT_EXIT;
}
if ((err = lseek(fd, 0, SEEK_SET)) < 0) {
rexec_err("Rexec pidmap file:%s lseek 0 failed errno:%d rexec exit", path, err);
- return -1;
+ return REXEC_EVENT_EXIT;
}
sprintf(buf, "%d", peerpid);
if ((err = write(fd, buf, strlen(buf))) <= 0) {
rexec_err("Rexec pidmap file:%s write pid:%d failed errno:%d rexec exit.", path, peerpid, err);
- return -1;
+ return REXEC_EVENT_EXIT;
+ }
+ if (g_rexec.rexec_hs_fd[PIPE_WRITE] != -1 && g_rexec.rexec_hs_fd[PIPE_READ] != -1) {
+ err = write(g_rexec.rexec_hs_fd[PIPE_WRITE], "1", 1);
+ if (err <= 0) {
+ rexec_err("rexec handshake write 1 failed, hs write:%d.", g_rexec.rexec_hs_fd[PIPE_WRITE]);
+ return REXEC_EVENT_ERR;
+ }
+ } else {
+ char msg[sizeof(struct rexec_msg) + 1];
+ struct rexec_msg *hs = msg;
+ char *ok = hs->msg;
+ hs->msgtype = REXEC_HANDSHAKE;
+ hs->msglen = 1;
+ *ok = '1';
+ if (write(evt->fd, hs, sizeof(struct rexec_msg) + 1) <= 0) {
+ rexec_err("send handshake failed, remote process will die");
+ return REXEC_EVENT_EXIT;
+ }
}
break;
}
@@ -159,6 +194,35 @@ static int rexec_conn_msg(int connfd, int *exit_status, int *pidfd)
rexec_log("Rexec conn recv msgtype:%d argc:%d stdno:%d msglen:%d",
head.msgtype, head.argc, head.stdno, head.msglen);
+ return REXEC_EVENT_OK;
+}
+
+static struct rexec_client_event *rexec_add_event(int efd, int fd, int outfd, int (*handler)(struct rexec_client_event *))
+{
+ struct rexec_client_event *event = (struct rexec_client_event *)malloc(sizeof(struct rexec_client_event));
+ if (event == NULL) {
+ rexec_err("malloc failed.");
+ return NULL;
+ }
+ event->fd = fd;
+ event->outfd = outfd;
+ event->handler = handler;
+ struct epoll_event evt;
+ evt.data.ptr = (void *)event;
+ evt.events = EPOLLIN;
+ if (-1 == epoll_ctl(efd, EPOLL_CTL_ADD, event->fd, &evt)) {
+ rexec_err("epoll ctl add fd:%d event failed.", event->fd);
+ free(event);
+ return NULL;
+ }
+ return event;
+}
+
+static int rexec_del_event(struct rexec_client_event *event)
+{
+ // close will del fd in epoll list
+ close(event->fd);
+ free(event);
return 0;
}
@@ -166,48 +230,48 @@ enum {
REPOL_IN_INDEX = 0,
REPOL_OUT_INDEX,
REPOL_ERR_INDEX,
- REPOL_CONN_INDEX,
REPOL_INV_INDEX,
};
-static int rexec_run(int rstdin, int rstdout, int rstderr, int connfd, char *argv[])
+static int rexec_std_event(int efd, int rstdin, int rstdout, int rstderr)
{
- int exit_status = EXIT_FAILURE;
-#define REXEC_MAX_EVENTS 4
- int infds[4] = {STDIN_FILENO, rstdout, rstderr, connfd};
- int outfds[4] = {rstdin, STDOUT_FILENO, STDERR_FILENO, connfd};
+ #define REXEC_MAX_EVENTS 4
+ int infds[REPOL_INV_INDEX] = {STDIN_FILENO, rstdout, rstderr};
+ int outfds[REPOL_INV_INDEX] = {rstdin, STDOUT_FILENO, STDERR_FILENO};
- int efd = epoll_create1(0);
- if (efd == -1) {
- rexec_err("epoll create1 failed, errno:%d.", errno);
- return exit_status;
- }
- struct epoll_event evt;
for (int i = 0; i < REPOL_INV_INDEX; i++) {
- evt.data.u32 = i;
- evt.events = EPOLLIN;
- if (-1 == epoll_ctl(efd, EPOLL_CTL_ADD, infds[i], &evt)) {
+ if (NULL == rexec_add_event(efd, infds[i], outfds[i], rexec_io)) {
rexec_err("epoll ctl add fd:%d event failed and ignore this mistake.", infds[i]);
continue;
} else {
if (rexec_set_nonblock(infds[i], 1) != 0) {
rexec_err("rexec set fd:%d i:%d non block failed.", infds[i], i);
- return exit_status;
+ return -1;
}
}
}
+ return 0;
+}
+
+static int rexec_run(int efd, int connfd, char *argv[])
+{
+ int pidfd = -1;
+ int exit_status = EXIT_FAILURE;
+
+ struct rexec_client_event *connevt = rexec_add_event(efd, connfd, -1, rexec_conn_msg);
+ if (NULL == connevt || rexec_set_nonblock(connfd, 1) != 0) {
+ // process will exit, fd or mem resource will free by kernel soon
+ rexec_err("rexec add connfd event failed");
+ return exit_status;
+ }
+ // 这两个指针只能在当前函数上下文使用，是当前函数栈指针
+ connevt->exit_status = &exit_status;
+ connevt->pidfd = &pidfd;
struct epoll_event *evts = calloc(REXEC_MAX_EVENTS, sizeof(struct epoll_event));
if (evts == NULL) {
rexec_err("init calloc evts failed.");
goto end;
}
- int buflen = REXEC_MSG_LEN;
- char *buf = (char *)malloc(buflen);
- int pidfd = -1;
- if (buf == NULL) {
- rexec_err("Rexec malloc failed.");
- goto free_end;
- }
rexec_log("Rexec process start run, as proxy of remote %s", argv[1]);
while (1) {
int n = epoll_wait(efd, evts, REXEC_MAX_EVENTS, 1000);
@@ -219,23 +283,16 @@ static int rexec_run(int rstdin, int rstdout, int rstderr, int connfd, char *arg
continue;
}
for (int i = 0; i < n; i++) {
- int infd = -1;
- int outfd = -1;
- if (evts[i].data.u32 >= REPOL_INV_INDEX) {
- rexec_err("invalid epoll events index data:%d", evts[i].data.u32);
- continue;
+ struct rexec_client_event *evt = (struct rexec_client_event *)evts[i].data.ptr;
+ int ret = evt->handler(evt);
+ if (evts[i].events & EPOLLHUP || ret == REXEC_EVENT_EXIT) {
+ process_exit = 1;
}
- infd = infds[evts[i].data.u32];
- outfd = outfds[evts[i].data.u32];
- if (infd == connfd) {
- if (evts[i].events & EPOLLHUP || rexec_conn_msg(connfd, &exit_status, &pidfd) == -1)
- process_exit = 1;
- } else {
- if (rexec_io(infd, outfd, buf, buflen) == -1) {
- close(infd);
- }
+ if (ret == REXEC_EVENT_DEL) {
+ rexec_del_event(evt);
}
}
+ // process will exit, and free all resource and exit
if (process_exit) {
rexec_log("Rexec process %s exit.", argv[1]);
break;
@@ -250,8 +307,6 @@ static int rexec_run(int rstdin, int rstdout, int rstderr, int connfd, char *arg
remove(path);
}
- free(buf);
-
free_end:
free(evts);
@@ -319,7 +374,7 @@ struct rexec_fdinfo {
int offset;
};
-static inline int rexec_is_reg_file(int fd)
+static inline unsigned int rexec_fd_mode(int fd)
{
struct stat st;
char path[32] = {0};
@@ -327,9 +382,13 @@ static inline int rexec_is_reg_file(int fd)
rexec_err("get fd:%d fstat failed, errno:%d", fd, errno);
return 0;
}
- if (S_ISREG(st.st_mode)) {
+ return st.st_mode;
+}
+
+static inline int rexec_is_reg_file(int fd)
+{
+ if (S_ISREG(rexec_fd_mode(fd)))
return 1;
- }
return 0;
}
@@ -429,16 +488,85 @@ err_end:
return NULL;
}
+static int rexec_handshake_proc(struct rexec_client_event *evt)
+{
+ char msg[sizeof(struct rexec_msg) + 1];
+ struct rexec_msg *hs = msg;
+ int ret = read(evt->fd, hs->msg, 1);
+ if (ret <= 0) {
+ rexec_err("read from handshake pipe failed, ret:%d err:%d", ret, errno);
+ return REXEC_EVENT_DEL;
+ }
+ hs->msgtype = REXEC_HANDSHAKE;
+ hs->msglen = 1;
+ ret = write(evt->outfd, hs, sizeof(struct rexec_msg) + 1);
+ if (ret < 0) {
+ rexec_err("send handshake failed, connfd:%d.", evt->outfd);
+ }
+ return REXEC_EVENT_OK;
+}
+
+static int rexec_handshake_init(int efd, int connfd)
+{
+ char *hs_read = getenv("REXEC_HANDSHAKE_RD");
+ char *hs_write = getenv("REXEC_HANDSHAKE_WR");
+
+ if (hs_read == NULL || hs_write == NULL) {
+ rexec_log("handshake not in effect, read:%lx write%lx", hs_read, hs_write);
+ return 0;
+ }
+ g_rexec.rexec_hs_fd[PIPE_READ] = atoi(hs_read);
+ g_rexec.rexec_hs_fd[PIPE_WRITE] = atoi(hs_write);
+ if (g_rexec.rexec_hs_fd[PIPE_READ] <= STDERR_FILENO || g_rexec.rexec_hs_fd[PIPE_WRITE] <= STDERR_FILENO) {
+ rexec_log("handshake invalid fd read:%d write:%d", g_rexec.rexec_hs_fd[PIPE_READ], g_rexec.rexec_hs_fd[PIPE_WRITE]);
+ goto err_end;
+ }
+ if (!S_ISFIFO(rexec_fd_mode(g_rexec.rexec_hs_fd[PIPE_READ])) || !S_ISFIFO(rexec_fd_mode(g_rexec.rexec_hs_fd[PIPE_WRITE]))) {
+ rexec_err("handshake fd mode not fifo:%d %d", g_rexec.rexec_hs_fd[PIPE_READ], g_rexec.rexec_hs_fd[PIPE_WRITE]);
+ goto err_end;
+ }
+ if (rexec_add_event(efd, g_rexec.rexec_hs_fd[PIPE_READ], connfd, rexec_handshake_proc) == NULL) {
+ rexec_err("add handshake pipe read fd:%d to epoll failed", g_rexec.rexec_hs_fd[PIPE_READ]);
+ goto err_end;
+ }
+ rexec_log("handshake effect read:%d write:%d", g_rexec.rexec_hs_fd[PIPE_READ], g_rexec.rexec_hs_fd[PIPE_WRITE]);
+ return 0;
+err_end:
+ g_rexec.rexec_hs_fd[PIPE_READ] = -1;
+ g_rexec.rexec_hs_fd[PIPE_WRITE] = -1;
+ return -1;
+}
+
+static void rexec_global_var_init()
+{
+ memset(&g_rexec, 0, sizeof(g_rexec));
+ g_rexec.rexec_hs_fd[PIPE_READ] = -1;
+ g_rexec.rexec_hs_fd[PIPE_WRITE] = -1;
+ return;
+}
+
int main(int argc, char *argv[])
{
rexec_log_init();
rexec_clear_pids();
+ int efd = epoll_create1(0);
+ if (efd == -1) {
+ rexec_err("epoll create1 failed, errno:%d.", errno);
+ return -1;
+ }
+ rexec_global_var_init();
+
int connfd = rexec_conn_to_server();
if (connfd < 0) {
rexec_err("Rexec connect to server failed, errno:%d", errno);
return -1;
}
+
+ if (rexec_handshake_init(efd, connfd) != 0) {
+ rexec_err("Rexec handshake environment set but get error.");
+ return -1;
+ }
rexec_log("Remote exec binary:%s", argv[1]);
int arglen = rexec_calc_argv_len(argc - 1, &argv[1]);
@@ -513,7 +641,11 @@ int main(int argc, char *argv[])
close(rstdin[0]);
close(rstdout[1]);
close(rstderr[1]);
- exit_status = rexec_run(rstdin[1], rstdout[0], rstderr[0], connfd, argv);
+ if (rexec_std_event(efd, rstdin[1], rstdout[0], rstderr[0]) != 0) {
+ rexec_err("add std event failed");
+ goto err_end;
+ }
+ exit_status = rexec_run(efd, connfd, argv);
close(rstdin[1]);
close(rstdout[0]);
close(rstderr[0]);
diff --git a/qtfs/rexec/rexec.h b/qtfs/rexec/rexec.h
index ba7c2be..ce1280a 100644
--- a/qtfs/rexec/rexec.h
+++ b/qtfs/rexec/rexec.h
@@ -24,6 +24,13 @@ enum {
PIPE_WRITE,
};
+enum {
+ REXEC_EVENT_OK,
+ REXEC_EVENT_DEL, // del this event
+ REXEC_EVENT_EXIT, // exit process
+ REXEC_EVENT_ERR,
+};
+
enum {
REXEC_STDIN = 0x5a,
REXEC_STDOUT,
@@ -45,6 +52,7 @@ enum rexec_msgtype {
REXEC_KILL, // kill process
REXEC_PIPE, // client send a pipefd as stdin/out/err to server
REXEC_PIDMAP, // server send remote process's pid to client
+ REXEC_HANDSHAKE,
};
struct rexec_msg {
diff --git a/qtfs/rexec/rexec_server.c b/qtfs/rexec/rexec_server.c
index 686c051..2aa3275 100644
--- a/qtfs/rexec/rexec_server.c
+++ b/qtfs/rexec/rexec_server.c
@@ -65,12 +65,6 @@ struct rexec_event {
int (*handler)(struct rexec_event *);
};
-enum {
- REXEC_EVENT_OK,
- REXEC_EVENT_ERR,
- REXEC_EVENT_DEL,
-};
-
static int rexec_add_event(int efd, int fd, int pid, int (*handler)(struct rexec_event *))
{
struct rexec_event *event = (struct rexec_event *)malloc(sizeof(struct rexec_event));
@@ -86,6 +80,7 @@ static int rexec_add_event(int efd, int fd, int pid, int (*handler)(struct rexec
evt.events = EPOLLIN;
if (-1 == epoll_ctl(efd, EPOLL_CTL_ADD, event->fd, &evt)) {
rexec_err("epoll ctl add fd:%d event failed.", event->fd);
+ free(event);
return -1;
}
return 0;
@@ -136,15 +131,6 @@ static int rexec_event_handshake(struct rexec_event *event)
rexec_log("Rexec recv son pid:%d, connfd:%d", sonpid, connfd);
rexec_hash_insert_direct(child_hash, sonpid, connfd);
-
- struct rexec_msg head;
- head.msgtype = REXEC_PIDMAP;
- head.msglen = 0;
- head.pid = sonpid;
- ret = write(connfd, &head, sizeof(struct rexec_msg));
- if (ret <= 0) {
- rexec_err("Rexec send son pid:%d to client failed, ret:%d errno:%d", sonpid, ret, errno);
- }
rexec_add_event(main_epoll_fd, connfd, sonpid, rexec_event_process_manage);
// 成功后同样要删除这个pipe监听事件，删除时会close掉fd
@@ -326,7 +312,7 @@ static int rexec_start_new_process(int newconnfd)
int scmfd = -1;
int len = sizeof(struct rexec_msg);
memset(&head, 0, sizeof(struct rexec_msg));
- int ret = rexec_recvmsg(newconnfd, (char *)&head, len, &scmfd, MSG_WAITALL);
+ ret = rexec_recvmsg(newconnfd, (char *)&head, len, &scmfd, MSG_WAITALL);
if (ret <= 0) {
rexec_log("recvmsg ret:%d, errno:%d", ret, errno);
goto err_to_parent;
@@ -375,14 +361,45 @@ static int rexec_start_new_process(int newconnfd)
goto err_free;
}
+ char *ack;
int mypid = getpid();
+ char msg[sizeof(struct rexec_msg) + 1];
+ struct rexec_msg *pm = msg;
+ pm->msgtype = REXEC_PIDMAP;
+ pm->msglen = 0;
+ pm->pid = mypid;
+ ret = write(newconnfd, pm, sizeof(struct rexec_msg));
+ if (ret <= 0) {
+ rexec_err("Rexec send son pid:%d to client failed, ret:%d errno:%d", mypid, ret, errno);
+ } else {
+retry:
+ rexec_log("Waiting for rexec client handshake...");
+ ret = read(newconnfd, pm, sizeof(struct rexec_msg) + 1);
+ if (ret <= 0) {
+ rexec_err("Recv handshake failed, ret:%d err:%d", ret, errno);
+ goto err_to_parent;
+ }
+ if (pm->msgtype != REXEC_HANDSHAKE) {
+ rexec_err("Recv unexpected msg:%d", pm->msgtype);
+ goto retry;
+ }
+ ack = pm->msg;
+ if (*ack != '1') {
+ rexec_err("recv error handshake ack from client:%c, exit now", *ack);
+ goto err_to_parent;
+ }
+ }
// 写会PID必须放在基于newconnfd接收完所有消息之后，
// 后面newconnfd的控制权交回父进程rexec server服务进程
- write(pipefd[PIPE_WRITE], &mypid, sizeof(int));
+ if (write(pipefd[PIPE_WRITE], &mypid, sizeof(int)) <= 0) {
+ rexec_err("write pid to parent failed, pipefd:%d.", pipefd[PIPE_WRITE]);
+ }
// 子进程不再使用pipe write和connfd
close(pipefd[PIPE_WRITE]);
close(newconnfd);
+ rexec_log("handshake over normaly, continue to exec new process:%s.", binary);
+
// rexec_shim_entry argv like:
// argv[0]: binary
// argv[1]: -f
--
2.33.0

171
0002-uds-add-vsock-mode.patch Normal file
View File

@ -0,0 +1,171 @@
From 9816a45d35ffa187b5830a417147afa9695dcead Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Sat, 3 Jun 2023 11:01:23 +0800
Subject: uds add vsock mode
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/ipc/Makefile | 7 +++++++
qtfs/ipc/uds_main.c | 43 ++++++++++++++++++++++++++++++++++++++-----
qtfs/ipc/uds_main.h | 12 +++++++++---
3 files changed, 54 insertions(+), 8 deletions(-)
diff --git a/qtfs/ipc/Makefile b/qtfs/ipc/Makefile
index 2136f4f..3b851f3 100644
--- a/qtfs/ipc/Makefile
+++ b/qtfs/ipc/Makefile
@@ -3,6 +3,9 @@ CFLAGS += -g -O2
CFLAGS += -fstack-protector-strong
CFLAGS += -fPIE -pie -fPIC
CFLAGS += -D_FORTIFY_SOURCE=2
+ifdef UDS_TEST_MODE
+CFLAGS += -DUDS_TEST_MODE
+endif
LDFLAGS += -s
LDFLAGS += -Wl,-z,now
LDFLAGS += -Wl,-z,noexecstack
@@ -12,6 +15,10 @@ all: udsproxyd libudsproxy.so
udsproxyd: uds_event.o uds_main.o
gcc $(LDFLAGS) -o udsproxyd $^ -I../ $(DEPGLIB)
+ @test -z $(UDS_TEST_MODE) || echo "Important risk warning: The test mode is turned on,\
+ and udsproxyd will expose the network port, which will bring security risks and is only for\
+ testing! If you do not understand the risks, please don't use or compile again without\
+ UDS_TEST_MODE."
uds_event.o:
cc $(CFLAGS) -c -o uds_event.o uds_event.c $(DEPGLIB)
diff --git a/qtfs/ipc/uds_main.c b/qtfs/ipc/uds_main.c
index 2af4f06..adf0936 100644
--- a/qtfs/ipc/uds_main.c
+++ b/qtfs/ipc/uds_main.c
@@ -37,6 +37,7 @@
#include <sys/resource.h>
#include <sys/prctl.h>
#include <sys/file.h>
+#include <linux/vm_sockets.h>
#include "comm.h"
#include "uds_main.h"
@@ -179,15 +180,24 @@ void uds_main_loop(int efd, struct uds_thread_arg *arg)
#define UDS_MAX_LISTEN_NUM 64
int uds_build_tcp_connection(struct uds_conn_arg *arg)
{
+ int family = AF_VSOCK;
if (arg->cs > UDS_SOCKET_SERVER) {
uds_err("cs type %d is error.", arg->cs);
return -1;
}
- struct sockaddr_in sock_addr = {
- .sin_family = AF_INET,
- };
- int sock_fd = socket(AF_INET, SOCK_STREAM, 0);
+#ifdef UDS_TEST_MODE
+ family = AF_INET;
+ struct sockaddr_in sock_addr;
+ memset(&sock_addr, 0, sizeof(sock_addr));
+ sock_addr.sin_family = AF_INET;
+#else
+ family = AF_VSOCK;
+ struct sockaddr_vm sock_addr;
+ memset(&sock_addr, 0, sizeof(sock_addr));
+ sock_addr.svm_family = AF_VSOCK;
+#endif
+ int sock_fd = socket(family, SOCK_STREAM, 0);
if (sock_fd < 0) {
uds_err("As %s failed, socket fd: %d, errno:%d.",
(arg->cs == UDS_SOCKET_CLIENT) ? "client" : "server",
@@ -197,8 +207,13 @@ int uds_build_tcp_connection(struct uds_conn_arg *arg)
arg->sockfd = sock_fd;
if (arg->cs == UDS_SOCKET_SERVER) {
+#ifdef UDS_TEST_MODE
sock_addr.sin_port = htons(p_uds_var->tcp.port);
sock_addr.sin_addr.s_addr = inet_addr(p_uds_var->tcp.addr);
+#else
+ sock_addr.svm_port = p_uds_var->vsock.port;
+ sock_addr.svm_cid = p_uds_var->vsock.cid;
+#endif
if (bind(sock_fd, (struct sockaddr *)&sock_addr, sizeof(sock_addr)) < 0) {
uds_err("As tcp server failed, bind error, errno:%d.",
errno);
@@ -209,13 +224,22 @@ int uds_build_tcp_connection(struct uds_conn_arg *arg)
goto close_and_return;
}
} else {
+#ifdef UDS_TEST_MODE
sock_addr.sin_port = htons(p_uds_var->tcp.peerport);
sock_addr.sin_addr.s_addr = inet_addr(p_uds_var->tcp.peeraddr);
- if (connect(arg->sockfd, (struct sockaddr *)&sock_addr, sizeof(struct sockaddr_in)) < 0) {
+#else
+ sock_addr.svm_port = p_uds_var->vsock.peerport;
+ sock_addr.svm_cid = p_uds_var->vsock.peercid;
+#endif
+ if (connect(arg->sockfd, (struct sockaddr *)&sock_addr, sizeof(sock_addr)) < 0) {
goto close_and_return;
}
arg->connfd = sock_fd;
+#ifdef UDS_TEST_MODE
uds_log("Connect to tcp server successed, ip:%s port:%u", p_uds_var->tcp.peeraddr, p_uds_var->tcp.peerport);
+#else
+ uds_log("Connect to vsock server successed, cid:%u port:%u", p_uds_var->vsock.peercid, p_uds_var->vsock.peerport);
+#endif
}
return 0;
@@ -599,6 +623,7 @@ static int uds_glob_var_init(char *argv[])
uds_err("work thread var malloc failed.");
return -1;
}
+#ifdef UDS_TEST_MODE
p_uds_var->tcp.port = atoi(argv[3]);
strncpy(p_uds_var->tcp.addr, argv[2], sizeof(p_uds_var->tcp.addr) - 1);
p_uds_var->tcp.peerport = atoi(argv[5]);
@@ -607,6 +632,14 @@ static int uds_glob_var_init(char *argv[])
uds_log("uds proxy param thread num:%d ip:%s port:%u peerip:%s port:%u",
p_uds_var->work_thread_num, p_uds_var->tcp.addr, p_uds_var->tcp.port,
p_uds_var->tcp.peeraddr, p_uds_var->tcp.peerport);
+#else
+ // vsock param: <thread num> <local cid> <local port> <peer cid> <peer port>
+ // port and peerport is checked before
+ p_uds_var->vsock.cid = atoi(argv[2]);
+ p_uds_var->vsock.port = myport;
+ p_uds_var->vsock.peercid = atoi(argv[4]);
+ p_uds_var->vsock.peerport = peerport;
+#endif
g_event_var = (struct uds_event_global_var *)malloc(sizeof(struct uds_event_global_var) * p_uds_var->work_thread_num);
if (g_event_var == NULL) {
free(p_uds_var->efd);
diff --git a/qtfs/ipc/uds_main.h b/qtfs/ipc/uds_main.h
index 3903ec8..de400f8 100644
--- a/qtfs/ipc/uds_main.h
+++ b/qtfs/ipc/uds_main.h
@@ -111,15 +111,21 @@ struct uds_global_var {
struct uds_thread_arg *work_thread;
int loglevel;
char **logstr;
+#ifdef UDS_TEST_MODE
struct _tcp {
char addr[20];
unsigned short port;
char peeraddr[20];
unsigned short peerport;
} tcp;
- struct _uds {
- char sun_path[UDS_SUN_PATH_LEN];
- } uds;
+#else
+ struct _vsock {
+ unsigned int cid;
+ unsigned int port;
+ unsigned int peercid;
+ unsigned int peerport;
+ } vsock;
+#endif
};
enum uds_cs {
UDS_SOCKET_CLIENT = 1,
--
2.33.0

26
0003-fix-a-rexec_server-bug-on-child-process.patch Normal file
View File

@ -0,0 +1,26 @@
From 754345d6dffc581e81670f4d1b18396003349a18 Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Mon, 5 Jun 2023 17:24:10 +0800
Subject: fix a rexec_server bug on child process
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/rexec/rexec_server.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/qtfs/rexec/rexec_server.c b/qtfs/rexec/rexec_server.c
index 2aa3275..bdc6ea4 100644
--- a/qtfs/rexec/rexec_server.c
+++ b/qtfs/rexec/rexec_server.c
@@ -431,7 +431,7 @@ err_to_parent:
write(pipefd[PIPE_WRITE], &errpid, sizeof(int));
} while (0);
- return ret;
+ exit(0);
}
// 道生一
--
2.33.0

24
0004-add-uds-test-mode-in-engine-compile.patch Normal file
View File

@ -0,0 +1,24 @@
From 5185a08ec742ee044c4efa7fa5e1ae9fb454f80f Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Mon, 5 Jun 2023 20:53:39 +0800
Subject: add uds test mode in engine compile
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/qtfs_server/Makefile | 1 +
1 file changed, 1 insertion(+)
diff --git a/qtfs/qtfs_server/Makefile b/qtfs/qtfs_server/Makefile
index 0a8ed39..d00ce66 100644
--- a/qtfs/qtfs_server/Makefile
+++ b/qtfs/qtfs_server/Makefile
@@ -1,5 +1,6 @@
ifdef QTFS_TEST_MODE
ccflags-y += -I$(src)/../ -I$(src) -I$(src)/../ipc/ -I$(src)/../include/ -DQTFS_SERVER -DQTFS_TEST_MODE
+CFLAGS += -DUDS_TEST_MODE
else
ccflags-y += -I$(src)/../ -I$(src) -I$(src)/../ipc/ -I$(src)/../include/ -DQTFS_SERVER
endif
--
2.33.0

69
0005-modify-qtfs_dentry_revalidate.patch Normal file
View File

@ -0,0 +1,69 @@
From 607c1e329ac107087fdadca0c997d3e0f66b3b06 Mon Sep 17 00:00:00 2001
From: yangxin <245051644@qq.com>
Date: Thu, 8 Jun 2023 06:54:09 +0000
Subject: modify qtfs_dentry_revalidate
Signed-off-by: yangxin <245051644@qq.com>
---
qtfs/qtfs/sb.c | 32 ++++++++++++++++++++++----------
1 file changed, 22 insertions(+), 10 deletions(-)
diff --git a/qtfs/qtfs/sb.c b/qtfs/qtfs/sb.c
index b6bb3d7..2038c55 100644
--- a/qtfs/qtfs/sb.c
+++ b/qtfs/qtfs/sb.c
@@ -1510,29 +1510,41 @@ const struct xattr_handler *qtfs_xattr_handlers[] = {
int qtfs_dentry_revalidate(struct dentry *dentry, unsigned int flags)
{
struct qtfs_conn_var_s *pvar = NULL;
- struct qtreq_mount *req = NULL;
- struct qtrsp_mount *rsp = NULL;
-
- // 1 means valid; 0 means invalid
- if (dentry && dentry->d_inode && S_ISDIR(dentry->d_inode->i_mode)) {
+ struct qtreq_getattr *req;
+ struct qtrsp_getattr *rsp;
+ struct inode *inode = dentry->d_inode;
+ if (dentry && dentry->d_inode) {
if (jiffies - dentry->d_time < 2000)
return 1;
-
pvar = qtfs_conn_get_param();
- if (!pvar)
+ if (!pvar) {
+ qtfs_err("Failed to get qtfs sock var\n");
return 0;
+ }
req = pvar->conn_ops->get_conn_msg_buf(pvar, QTFS_SEND);
qtfs_fullname(req->path, dentry, PATH_MAX);
- rsp = qtfs_remote_run(pvar, QTFS_REQ_MOUNT, strlen(req->path));
- if (IS_ERR_OR_NULL(rsp) || rsp->ret != QTFS_OK) {
+ req->request_mask = STATX_BASIC_STATS;
+ req->query_flags = 0;
+
+ rsp = qtfs_remote_run(pvar, QTFS_REQ_GETATTR, QTFS_SEND_SIZE(struct qtreq_getattr, req->path));
+ if (IS_ERR_OR_NULL(rsp)) {
+ qtfs_conn_put_param(pvar);
+ return 0;
+ }
+ if (rsp->ret) {
qtfs_conn_put_param(pvar);
return 0;
}
+ if (!inode || inode->i_ino != rsp->stat.ino || inode->i_mode != rsp->stat.mode) {
+ if (inode->i_nlink > 0)
+ drop_nlink(inode);
+ qtfs_conn_put_param(pvar);
+ return 0;
+ }
qtfs_conn_put_param(pvar);
dentry->d_time = jiffies;
- return 1;
}
return 1;
}
--
2.33.0

55
0006-add-cmake-compile-for-user-binary-and-so.patch Normal file
View File

@ -0,0 +1,55 @@
From fbde8e2d0b6f66daf7aacb7348dffb97721bca7c Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Thu, 8 Jun 2023 15:58:09 +0800
Subject: add cmake compile for user binary and so
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/CMakeLists.txt | 34 ++++++++++++++++++++++++++++++++++
1 file changed, 34 insertions(+)
create mode 100644 qtfs/CMakeLists.txt
diff --git a/qtfs/CMakeLists.txt b/qtfs/CMakeLists.txt
new file mode 100644
index 0000000..d5b65ec
--- /dev/null
+++ b/qtfs/CMakeLists.txt
@@ -0,0 +1,34 @@
+cmake_minimum_required(VERSION 3.0.0)
+
+project(qtfs)
+
+set(CMAKE_C_FLAGS "-g -O2 -fstack-protector-strong -fPIE -pie -fPIC -D_FORTIFY_SOURCE=2 -s -Wl,-z,now -Wl,-z,noexecstack")
+
+# Build rexec and rexec_server
+add_executable(rexec rexec/rexec.c rexec/rexec_sock.c)
+add_executable(rexec_server rexec/rexec_server.c rexec/rexec_sock.c rexec/rexec_shim.c)
+target_include_directories(rexec_server PRIVATE /usr/include/glib-2.0 /usr/lib64/glib-2.0/include)
+target_link_libraries(rexec PRIVATE json-c)
+target_link_libraries(rexec_server PRIVATE json-c glib-2.0)
+
+# Build udsproxyd and libudsproxy.so
+add_executable(udsproxyd ipc/uds_event.c ipc/uds_main.c)
+add_library(udsproxy SHARED ipc/uds_connector.c)
+target_include_directories(udsproxyd PRIVATE include/ /usr/include/glib-2.0 /usr/lib64/glib-2.0/include)
+target_link_libraries(udsproxyd PRIVATE pthread glib-2.0)
+
+# Build engine
+add_executable(engine ipc/uds_main.c ipc/uds_event.c qtfs_common/user_engine.c)
+target_include_directories(engine PRIVATE include/ ./ ipc/ /usr/include/glib-2.0 /usr/lib64/glib-2.0/include)
+target_link_libraries(engine PRIVATE glib-2.0 pthread)
+target_compile_options(engine PRIVATE "-DQTFS_SERVER")
+
+if(DEFINED UDS_TEST_MODE OR DEFINED QTFS_TEST_MODE)
+target_compile_options(engine PRIVATE "-DUDS_TEST_MODE")
+target_compile_options(udsproxyd PRIVATE "-DUDS_TEST_MODE")
+message(WARNING "Important risk warning: the test mode is turned on, and qtfs will expose the network port, \
+ which will bring security risks and is only for testing! If you do not understand the risks,\
+ please don't use or compile again without test mode macro!")
+endif()
+
+set(ignoreMe "${QTFS_TEST_MODE}${UDS_TEST_MODE}")
\ No newline at end of file
--
2.33.0

150
0007-uds-remove-MSG_WAITALL-flags-to-timeout.patch Normal file
View File

@ -0,0 +1,150 @@
From 596a9553d90516375de169acdd513e4df26eab9d Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Thu, 8 Jun 2023 15:27:06 +0800
Subject: uds remove MSG_WAITALL flags to timeout
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/ipc/uds_event.c | 16 ++++++++--------
qtfs/ipc/uds_main.c | 36 ++++++++++++++++++++++++++++++++++++
qtfs/ipc/uds_main.h | 1 +
3 files changed, 45 insertions(+), 8 deletions(-)
diff --git a/qtfs/ipc/uds_event.c b/qtfs/ipc/uds_event.c
index b046b69..047949b 100644
--- a/qtfs/ipc/uds_event.c
+++ b/qtfs/ipc/uds_event.c
@@ -214,7 +214,7 @@ int uds_event_build_step2(void *arg, int epfd, struct uds_event_global_var *p_ev
struct uds_proxy_remote_conn_rsp rsp;
int len;
memset(buf, 0, sizeof(buf));
- len = recv(evt->fd, msg, sizeof(struct uds_proxy_remote_conn_req), MSG_WAITALL);
+ len = uds_recv_with_timeout(evt->fd, msg, sizeof(struct uds_proxy_remote_conn_req));
if (len == 0) {
uds_err("recv err msg:%d errno:%d", len, errno);
return EVENT_DEL;
@@ -390,7 +390,7 @@ int uds_event_tcp_listener(void *arg, int epfd, struct uds_event_global_var *p_e
int uds_build_connect2uds(struct uds_event *evt, struct uds_proxy_remote_conn_req *msg)
{
struct uds_conn_arg targ;
- int len = recv(evt->fd, msg, sizeof(struct uds_proxy_remote_conn_req), MSG_WAITALL);
+ int len = uds_recv_with_timeout(evt->fd, msg, sizeof(struct uds_proxy_remote_conn_req));
if (len <= 0) {
uds_err("recv failed, len:%d errno:%d", len, errno);
return EVENT_ERR;
@@ -441,7 +441,7 @@ err_ack:
int uds_build_pipe_proxy(int efd, struct uds_event *evt, struct uds_stru_scm_pipe *msg)
{
- int len = recv(evt->fd, msg, sizeof(struct uds_stru_scm_pipe), MSG_WAITALL);
+ int len = uds_recv_with_timeout(evt->fd, msg, sizeof(struct uds_stru_scm_pipe));
if (len <= 0) {
uds_err("recv failed, len:%d errno:%d", len, errno);
return EVENT_ERR;
@@ -477,7 +477,7 @@ int uds_event_remote_build(void *arg, int epfd, struct uds_event_global_var *p_e
int len;
int ret = EVENT_OK;
memset(p_event_var->iov_base, 0, p_event_var->iov_len);
- len = recv(evt->fd, bdmsg, sizeof(struct uds_tcp2tcp), MSG_WAITALL);
+ len = uds_recv_with_timeout(evt->fd, bdmsg, sizeof(struct uds_tcp2tcp));
if (len <= 0) {
uds_err("read no msg from sock:%d, len:%d", evt->fd, len);
return EVENT_DEL;
@@ -711,7 +711,7 @@ int uds_msg_tcp2uds_scm_pipe(struct uds_tcp2tcp *p_msg, struct uds_event *evt, i
int scmfd;
int fd[SCM_PIPE_NUM];
struct uds_stru_scm_pipe *p_pipe = (struct uds_stru_scm_pipe *)p_msg->data;
- int len = recv(evt->fd, p_pipe, p_msg->msglen, MSG_WAITALL);
+ int len = uds_recv_with_timeout(evt->fd, p_pipe, p_msg->msglen);
if (len <= 0) {
uds_err("recv data failed, len:%d", len);
return EVENT_DEL;
@@ -913,7 +913,7 @@ int uds_event_tcp2uds(void *arg, int epfd, struct uds_event_global_var *p_event_
msg.msg_controllen = p_event_var->msg_controlsendlen;
while (1) {
- int len = recv(evt->fd, p_msg, sizeof(struct uds_tcp2tcp), MSG_WAITALL);
+ int len = uds_recv_with_timeout(evt->fd, p_msg, sizeof(struct uds_tcp2tcp));
if (len <= 0) {
uds_err("recv no msg maybe sock is closed, delete this tcp2uds event, len:%d.", len);
goto close_event;
@@ -932,7 +932,7 @@ int uds_event_tcp2uds(void *arg, int epfd, struct uds_event_global_var *p_event_
uds_err("normal msg repeat recv fd:%d", evt->fd);
goto err;
}
- normal_msg_len = recv(evt->fd, p_event_var->iov_base_send, p_msg->msglen, MSG_WAITALL);
+ normal_msg_len = uds_recv_with_timeout(evt->fd, p_event_var->iov_base_send, p_msg->msglen);
if (normal_msg_len <= 0) {
uds_err("recv msg error:%d fd:%d", len, evt->fd);
goto close_event;
@@ -950,7 +950,7 @@ int uds_event_tcp2uds(void *arg, int epfd, struct uds_event_global_var *p_event_
}
memset(p_scm->path, 0, sizeof(p_scm->path));
// SCM RIGHTS msg proc
- len = recv(evt->fd, p_msg->data, p_msg->msglen, MSG_WAITALL);
+ len = uds_recv_with_timeout(evt->fd, p_msg->data, p_msg->msglen);
if (len <= 0) {
uds_err("recv data failed len:%d", p_msg->msglen);
return EVENT_DEL;
diff --git a/qtfs/ipc/uds_main.c b/qtfs/ipc/uds_main.c
index adf0936..65ae81d 100644
--- a/qtfs/ipc/uds_main.c
+++ b/qtfs/ipc/uds_main.c
@@ -88,6 +88,42 @@ int uds_event_delete(int efd, int fd)
return 0;
}
+int uds_recv_with_timeout(int fd, char *msg, int len)
+{
+#define TMOUT_BLOCK_SIZE 1024
+#define TMOUT_UNIT_MS 20
+#define TMOUT_INTERVAL 1
+#define TMOUT_MAX_MS 1000
+ int total_recv = 0;
+ int ret;
+ int tmout_ms = ((len / TMOUT_BLOCK_SIZE) + 1) * TMOUT_UNIT_MS;
+ if (len <= 0 || msg == NULL || fd < 0) {
+ uds_err("invalid param fd:%d len:%d or %s", fd, len, (msg == NULL) ? "msg is NULL" : "msg is not NULL");
+ return 0;
+ }
+ if (tmout_ms > TMOUT_MAX_MS)
+ tmout_ms = TMOUT_MAX_MS;
+ do {
+ ret = recv(fd, &msg[total_recv], len - total_recv, 0);
+ if (ret < 0) {
+ uds_err("recv failed ret:%d errno:%d", ret, errno);
+ return ret;
+ }
+ total_recv += ret;
+ if (total_recv > len) {
+ uds_err("fatal error total recv:%d longger than target len:%d", total_recv, len);
+ return 0;
+ }
+ if (total_recv == len) {
+ return total_recv;
+ }
+ usleep(TMOUT_INTERVAL * 1000);
+ tmout_ms -= TMOUT_INTERVAL;
+ } while (tmout_ms > 0);
+ uds_err("Fatal error, the target recv len:%d and only %d length is received when it time out", len, total_recv);
+ return 0;
+}
+
#pragma GCC diagnostic ignored "-Wpointer-to-int-cast"
int uds_event_tmout_item(gpointer key, gpointer value, gpointer data)
{
diff --git a/qtfs/ipc/uds_main.h b/qtfs/ipc/uds_main.h
index de400f8..f69e4b3 100644
--- a/qtfs/ipc/uds_main.h
+++ b/qtfs/ipc/uds_main.h
@@ -152,6 +152,7 @@ int uds_event_insert(int efd, struct uds_event *event);
int uds_hash_insert_dirct(GHashTable *table, int key, struct uds_event *value);
void *uds_hash_lookup_dirct(GHashTable *table, int key);
int uds_hash_remove_dirct(GHashTable *table, int key);
+int uds_recv_with_timeout(int fd, char *msg, int len);
#ifdef QTFS_SERVER
int uds_proxy_main(int argc, char *argv[]);
--
2.33.0

35
0008-Modify-the-logic-of-rexec-proxy-process-exit-by-mist.patch Normal file
View File

@ -0,0 +1,35 @@
From 802188d8d9a06a0c963ecee0b7d30b379dfe2814 Mon Sep 17 00:00:00 2001
From: yangxin <245051644@qq.com>
Date: Thu, 8 Jun 2023 12:22:58 +0000
Subject: Modify the logic of rexec proxy process exit by mistake
Signed-off-by: yangxin <245051644@qq.com>
---
qtfs/rexec/rexec.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/qtfs/rexec/rexec.c b/qtfs/rexec/rexec.c
index 489ebec..5439fac 100644
--- a/qtfs/rexec/rexec.c
+++ b/qtfs/rexec/rexec.c
@@ -109,7 +109,7 @@ static int rexec_io(struct rexec_client_event *evt)
ret = write(evt->outfd, buf, len);
if (ret <= 0) {
rexec_err("Read from fd:%d len:%d write to fd:%d failed ret:%d", evt->fd, len, evt->outfd, ret);
- return REXEC_EVENT_EXIT;
+ return REXEC_EVENT_DEL;
}
if (ret != len) {
rexec_err("Read from fd:%d len:%d but write to fd:%d ret:%d", evt->fd, len, evt->outfd, ret);
@@ -285,7 +285,7 @@ static int rexec_run(int efd, int connfd, char *argv[])
for (int i = 0; i < n; i++) {
struct rexec_client_event *evt = (struct rexec_client_event *)evts[i].data.ptr;
int ret = evt->handler(evt);
- if (evts[i].events & EPOLLHUP || ret == REXEC_EVENT_EXIT) {
+ if (ret == REXEC_EVENT_EXIT) {
process_exit = 1;
}
if (ret == REXEC_EVENT_DEL) {
--
2.33.0

325
0009-Keep-doc-same-with-docs-in-openEuler.patch Normal file
View File

@ -0,0 +1,325 @@
From 67b7259d810c9e07227e7a1475acdc1141e4424e Mon Sep 17 00:00:00 2001
From: Weifeng Su <suweifeng1@huawei.com>
Date: Thu, 8 Jun 2023 14:17:00 +0000
Subject: Keep doc same with docs in openEuler
Signed-off-by: Weifeng Su <suweifeng1@huawei.com>
---
qtfs/doc/overview.md | 11 ++
...77\347\224\250\346\211\213\345\206\214.md" | 69 ++++++++
...40\346\204\237\345\215\270\350\275\275.md" | 31 ++++
...50\347\275\262\346\214\207\345\257\274.md" | 166 ++++++++++++++++++
7 files changed, 277 insertions(+)
create mode 100644 qtfs/doc/overview.md
create mode 100644 "qtfs/doc/qtfs\345\205\261\344\272\253\346\226\207\344\273\266\347\263\273\347\273\237\346\236\266\346\236\204\345\217\212\344\275\277\347\224\250\346\211\213\345\206\214.md"
create mode 100644 "qtfs/doc/\345\256\271\345\231\250\347\256\241\347\220\206\351\235\242\346\227\240\346\204\237\345\215\270\350\275\275.md"
create mode 100644 "qtfs/doc/\346\227\240\346\204\237\345\215\270\350\275\275\351\203\250\347\275\262\346\214\207\345\257\274.md"
diff --git a/qtfs/doc/overview.md b/qtfs/doc/overview.md
new file mode 100644
index 0000000..518deb0
--- /dev/null
+++ b/qtfs/doc/overview.md
@@ -0,0 +1,11 @@
+# 容器管理面DPU无感卸载指南
+
+本文档介绍基于openEuler操作系统的容器管理面DPU无感卸载功能特性及安装部署方法，该特性可以通过操作系统提供的统一抽象层，屏蔽容器管理面跨主机资源访问的差异，实现容器管理面业务无感卸载到DPU上。
+
+本文档适用于使用openEuler系统并希望了解和使用操作系统内核及容器的社区开发者、开源爱好者以及相关合作伙伴。使用人员需要具备以下经验和技能：
+
+- 熟悉Linux基本操作
+
+- 熟悉linux内核文件系统相关基础机制
+
+- 对kubernetes和docker有一定了解，熟悉docker及kubernetes部署及使用
\ No newline at end of file
diff --git "a/qtfs/doc/qtfs\345\205\261\344\272\253\346\226\207\344\273\266\347\263\273\347\273\237\346\236\266\346\236\204\345\217\212\344\275\277\347\224\250\346\211\213\345\206\214.md" "b/qtfs/doc/qtfs\345\205\261\344\272\253\346\226\207\344\273\266\347\263\273\347\273\237\346\236\266\346\236\204\345\217\212\344\275\277\347\224\250\346\211\213\345\206\214.md"
new file mode 100644
index 0000000..8088f48
--- /dev/null
+++ "b/qtfs/doc/qtfs\345\205\261\344\272\253\346\226\207\344\273\266\347\263\273\347\273\237\346\236\266\346\236\204\345\217\212\344\275\277\347\224\250\346\211\213\345\206\214.md"
@@ -0,0 +1,69 @@
+# qtfs
+
+## 介绍
+
+qtfs是一个共享文件系统项目，可部署在host-dpu的硬件架构上，也可以部署在2台服务器之间。以客户端服务器的模式工作，使客户端能通过qtfs访问服务端的指定文件系统，得到本地文件访问一致的体验。
+
+qtfs的特性：
+
++ 支持挂载点传播；
+
++ 支持proc、sys、cgroup等特殊文件系统的共享；
+
++ 支持远程文件读写的共享；
+
++ 支持在客户端对服务端的文件系统进行远程挂载；
+
++ 支持特殊文件的定制化处理；
+
++ 支持远端fifo、unix-socket等，并且支持epoll，使客户端和服务端像本地通信一样使用这些文件；
+
++ 支持基于host-dpu架构通过PCIe协议底层通信，性能大大优于网络；
+
++ 支持内核模块形式开发，无需对内核进行侵入式修改。
+
+## 软件架构
+
+软件大体框架图：
+
+![qtfs-arch](./figures/qtfs-arch.png)
+
+## 安装教程
+
+目录说明：
+
++ **qtfs**: 客户端内核模块相关代码，直接在该目录下编译客户端ko。
+
++ **qtfs_server**: 服务端内核模块相关代码，直接在该目录下编译服务端ko和相关程序。
+
++ **qtinfo**: 诊断工具，支持查询文件系统的工作状态以及修改log级别等。
+
++ **demo**、**test**、**doc**: 测试程序、演示程序以及项目资料等。
+
++ 根目录: 客户端与服务端通用的公共模块代码。
+
+首先找两台服务器（或虚拟机）配置内核编译环境：
+
+ 1. 要求内核版本在5.10或更高版本。
+  2. 安装内核开发包：yum install kernel-devel。
+
+服务端安装：
+
+ 1. cd qtfs_server
+ 2. make clean && make
+ 3. insmod qtfs_server.ko qtfs_server_ip=x.x.x.x qtfs_server_port=12345 qtfs_log_level=WARN
+ 4. ./engine 4096 16
+
+客户端安装：
+
+ 1. cd qtfs
+ 2. make clean && make
+ 3. insmod qtfs.ko qtfs_server_ip=x.x.x.x qtfs_server_port=12345 qtfs_log_level=WARN
+
+## 使用说明
+
+安装完成后，客户端通过挂载把服务端的文件系统让客户端可见，例如：
+
+ mount -t qtfs / /root/mnt/
+
+客户端进入"/root/mnt"后便可查看到server端的所有文件，以及对其进行相关操作。
diff --git "a/qtfs/doc/\345\256\271\345\231\250\347\256\241\347\220\206\351\235\242\346\227\240\346\204\237\345\215\270\350\275\275.md" "b/qtfs/doc/\345\256\271\345\231\250\347\256\241\347\220\206\351\235\242\346\227\240\346\204\237\345\215\270\350\275\275.md"
new file mode 100644
index 0000000..2e4be2f
--- /dev/null
+++ "b/qtfs/doc/\345\256\271\345\231\250\347\256\241\347\220\206\351\235\242\346\227\240\346\204\237\345\215\270\350\275\275.md"
@@ -0,0 +1,31 @@
+# 容器管理面无感卸载介绍
+
+## 概述
+
+在数据中心及云场景下，随着摩尔定律失效，通用处理单元CPU算力增长速率放缓，而同时网络IO类速率及性能不断攀升，二者增长速率差异形成的剪刀差，即当前通用处理器的处理能力无法跟上网络、磁盘等IO处理的需求。传统数据中心下越来越多的通用CPU算力被IO及管理面等占用，这部分资源损耗称之为数据中心税（Data-center Tax）。据AWS统计，数据中心税可能占据数据中心算力的30%以上，部分场景下甚至可能更多。
+
+DPU的出现就是为了将这部分算力资源从主机CPU上解放出来，通过将管理面、网络、存储、安全等能力卸载到专有的处理器芯片（DPU）上进行处理加速，达成降本增效的结果。目前主流云厂商如AWS、阿里云、华为云都通过自研芯片完成管理面及相关数据面的卸载，达成数据中心计算资源100%售卖给客户。
+
+管理面进程卸载到DPU可以通过对组件源码进行拆分达成，将源码根据功能逻辑拆分成独立运行的两部分，分别运行在主机和DPU，达成组件卸载的目的。但是这种做法有以下问题：一是影响组件的软件兼容性，组件后续版本升级和维护需要自己维护相关patch，带来一定的维护工作量；二是卸载工作无法被其他组件继承，后续组件卸载后仍需要进行代码逻辑分析和拆分等工作。为解决上述问题，本方案提出DPU的无感卸载，通过OS提供的抽象层，屏蔽应用在主机和DPU间跨主机访问的差异，让业务进程近似0改动达成卸载到DPU运行的目标，且这部分工作属于操作系统通用层，与上层业务无关，其他业务进行DPU卸载时也可以继承。
+
+## 架构介绍
+
+#### 容器管理面DPU无感卸载架构
+
+**图1**容器管理面DPU无感卸载架构
+
+![offload-arch](./figures/offload-arch.png)
+
+如图1所示，容器管理面卸载后，dockerd、kubelet等管理进程运行在DPU侧，容器进程本身运行在HOST，进程之间的交互关系由系统层提供对应的能力来保证：
+
+* 通信层：DPU和主机之间可能通过PCIe或网络进行通信，需要基于底层物理连接提供通信接口层，为上层业务提供通信接口。
+
+* 内核共享文件系统qtfs：容器管理面组件kubelet、dockerd与容器进程之间的主要交互通过文件系统进行；管理面工具需要为容器进程准备rootfs、volume等数据面路径；还需要在运行时通过proc文件系统、cgroup文件系统等控制和监控容器进程的资源及状态。共享文件系统的详细介绍参考[共享文件系统介绍](qtfs共享文件系统架构及使用手册.md)
+
+* 用户态卸载环境：用户态需要使用qtfs为容器管理面准备卸载后的运行时环境，将主机的容器管理及运行时相关目录远程挂载到DPU；另外由于需要挂载proc、sys、cgroup等系统管理文件系统，为防止对DPU原生系统功能的破坏，上述挂载动作都在chroot环境内完成。另外管理面（运行于DPU）和容器进程（运行于主机）之间仍存在调用关系，需要通过远程二进制执行工具（rexec）提供对应功能。
+
+容器管理面无感卸载的操作步骤可参考[部署指导文档](./无感卸载部署指导.md)
+
+> ![](./public_sys-resources/icon-note.gif)**说明**：
+>
+> 上述操作指导涉及对容器管理面组件的少量改动和rexec工具修改，这些修改基于指定版本，其他版本可基于实际执行环境做适配修改。文档中提供的patch仅供验证指导使用，不具备实际商用的条件
\ No newline at end of file
diff --git "a/qtfs/doc/\346\227\240\346\204\237\345\215\270\350\275\275\351\203\250\347\275\262\346\214\207\345\257\274.md" "b/qtfs/doc/\346\227\240\346\204\237\345\215\270\350\275\275\351\203\250\347\275\262\346\214\207\345\257\274.md"
new file mode 100644
index 0000000..c15eed9
--- /dev/null
+++ "b/qtfs/doc/\346\227\240\346\204\237\345\215\270\350\275\275\351\203\250\347\275\262\346\214\207\345\257\274.md"
@@ -0,0 +1,166 @@
+
+# 容器管理面无感卸载部署指导
+
+> ![](./public_sys-resources/icon-note.gif)**说明**：
+>
+> 本指导涉及对容器管理面组件的少量改动和rexec工具修改，这些修改基于指定版本，其他版本可基于实际执行环境做适配修改。文档中提供的patch仅供验证指导使用，不具备实际商用的条件。
+
+> ![](./public_sys-resources/icon-note.gif)**说明**：
+>
+> 当前共享文件系统之间通信通过网络完成，可通过网络互连的两台物理机器或VM模拟验证。
+>
+> 建议用户验证前先搭建可正常使用的kubernetes集群和容器运行环境，针对其中单个节点的管理面进程进行卸载验证，卸载环境（DPU）可选择一台具备网络连接的物理机或VM。
+
+## 简介
+
+容器管理面，即kubernetes、dockerd、containerd、isulad等容器的管理工具，而容器管理面卸载，即是将容器管理面卸载到与容器所在机器（以下称为HOST）之外的另一台机器（当前场景下是指DPU，一个具备独立运行环境的硬件集合）上运行。
+
+我们使用共享文件系统qtfs将HOST上与容器运行相关的目录挂载到DPU上，使得容器管理面工具（运行在DPU）可以访问到这些目录，并为容器（运行在HOST）准备运行所需要的环境，此处，因为需要挂载远端的proc和sys等特殊文件系统，所以，我们创建了一个专门的rootfs以作为kubernetes、dockerd的运行环境（以下称为`/another_rootfs`）。
+
+并且通过rexec执行容器的拉起、删除等操作，使得可以将容器管理面和容器分离在不同的两台机器上，远程对容器进行管理。
+
+## 相关组件补丁介绍
+
+#### rexec介绍
+
+rexec是一个用go语言开发的远程执行工具，基于docker/libchan下的[rexec](https://github.com/docker/libchan/tree/master/examples/rexec)示例工具改造而成，实现远程调用远端二进制的功能，为方便使用在rexec中增加了环境变量传递和监控原进程退出等能力。
+
+rexec工具的具体使用方式为在服务器端用`CMD_NET_ADDR=tcp://0.0.0.0:<端口号> rexec_server`的方式拉起rexec服务进程，然后在客户端用`CMD_NET_ADDR=tcp://<服务端ip>:<端口号> rexec [要执行的指令] `的方式启动，便可以调用rexec_server执行需要执行的指令，并等待指令执行结果返回。
+
+#### dockerd相关改动介绍
+
+对dockerd的改动基于18.09版本。
+
+在containerd中，暂时注释掉了通过hook调用libnetwork-setkey的部分，此处不影响容器的拉起。并且，为了docker load的正常使用，注释掉了在mounter_linux.go 中mount函数中一处错误的返回。
+
+最后，因为在容器管理面的运行环境中，将`/proc`挂在了服务端的proc文件系统，而本地的proc文件系统则挂载在了`/local_proc`，所以，dockerd以及containerd中的对`/proc/self/xxx`或者`/proc/getpid()/xxx`或者相关的文件系统访问的部分，我们统统将`/proc`改为了`/local_proc`。
+
+#### containerd相关改动介绍
+
+对于containerd的改动基于containerd-1.2-rc.1版本。
+
+在获取mountinfo时，因为`/proc/self/mountinfo`只能获取到dockerd本身在本地的mountinfo，而无法获取到服务端的mountinfo，所以，将其改为了`/proc/1/mountinfo`，使其通过获取服务端1号进程mountinfo的方式得到服务端的mountinfo。
+
+在contaienrd-shim中，将与containerd通信的unix socket改为了用tcp通信，containerd通过`SHIM_HOST`环境变量获取containerd-shim所运行环境的ip，即服务端ip。用shim的哈希值计算出一个端口号，并以此作为通信的端口，来拉起containerd-shim.
+
+并且，将原来的通过系统调用给contaienr-shim发信号的方式，改为了通过远程调用kill指令的方式向shim发信号，确保了docker杀死容器的行为可以正确的执行。
+
+#### kubernetes相关改动介绍
+
+kubelet暂不需要功能性改动，可能会遇到容器QoS管理器首次设置失败的错误，该错误不影响后续Pods拉起流程，暂时忽略该报错。
+
+## 容器管理面卸载操作指南
+
+在服务器端和客户端，都要拉起rexec_server。服务器端拉起rexec_server，主要是用于客户端创建容器时用rexec拉起containerd-shim，而客户端拉起rexec_server，则是为了执行containerd-shim对dockerd和containerd的调用。
+
+#### 服务器端
+
+创建容器管理面所需要的文件夹，然后插入qtfs_server.ko，并拉起engine进程。
+
+此外在服务器端，还需要创建rexec脚本/usr/bin/dockerd.
+
+``` shell
+#!/bin/bash
+CMD_NET_ADDR=tcp://<客户端ip>:<rexec端口号> rexec /usr/bin/dockerd $*
+```
+
+#### 客户端
+
+需要准备一个rootfs，作为dockerd与containerd的运行环境，通过如下的脚本，将dockerd、containerd所需要的服务端目录挂载到客户端。并且，需要确保在以下脚本中被挂载的远程目录在服务端和客户端都存在。
+
+``` shell
+#!/bin/bash
+mkdir -p /another_rootfs/var/run/docker/containerd
+iptables -t nat -N DOCKER
+echo "---------insmod qtfs ko----------"
+insmod /YOUR/QTFS/PATH/qtfs.ko qtfs_server_ip=<服务端ip> qtfs_log_level=INFO
+
+# chroot环境内的proc使用DPU的proc共享文件系统替换，需要将本机真实proc文件系统挂载到local_proc下使用
+mount -t proc proc /another_rootfs/local_proc/
+
+# 将chroot内环境与外部环境bind，方便进行配置和运行
+mount --bind /var/run/ /another_rootfs/var/run/
+mount --bind /var/lib/ /another_rootfs/var/lib/
+mount --bind /etc /another_rootfs/etc
+
+mkdir -p /another_rootfs/var/lib/isulad
+
+# 在chroot环境内创建并挂载dev、sys和cgroup文件系统
+mount -t devtmpfs devtmpfs /another_rootfs/dev/
+mount -t sysfs sysfs /another_rootfs/sys
+mkdir -p /another_rootfs/sys/fs/cgroup
+mount -t tmpfs tmpfs /another_rootfs/sys/fs/cgroup
+list="perf_event freezer files net_cls,net_prio hugetlb pids rdma cpu,cpuacct memory devices blkio cpuset"
+for i in $list
+do
+ echo $i
+ mkdir -p /another_rootfs/sys/fs/cgroup/$i
+ mount -t cgroup cgroup -o rw,nosuid,nodev,noexec,relatime,$i /another_rootfs/sys/fs/cgroup/$i
+done
+
+## common system dir
+mount -t qtfs -o proc /proc /another_rootfs/proc
+echo "proc"
+mount -t qtfs /sys /another_rootfs/sys
+echo "cgroup"
+
+# 挂载容器管理面所需要的共享目录
+mount -t qtfs /var/lib/docker/containers /another_rootfs/var/lib/docker/containers
+mount -t qtfs /var/lib/docker/containerd /another_rootfs/var/lib/docker/containerd
+mount -t qtfs /var/lib/docker/overlay2 /another_rootfs/var/lib/docker/overlay2
+mount -t qtfs /var/lib/docker/image /another_rootfs/var/lib/docker/image
+mount -t qtfs /var/lib/docker/tmp /another_rootfs/var/lib/docker/tmp
+mkdir -p /another_rootfs/run/containerd/io.containerd.runtime.v1.linux/
+mount -t qtfs /run/containerd/io.containerd.runtime.v1.linux/ /another_rootfs/run/containerd/io.containerd.runtime.v1.linux/
+mkdir -p /another_rootfs/var/run/docker/containerd
+mount -t qtfs /var/run/docker/containerd /another_rootfs/var/run/docker/containerd
+mount -t qtfs /var/lib/kubelet/pods /another_rootfs/var/lib/kubelet/pods
+```
+
+在/another_rootfs中，需要创建以下脚本，用来支持部分跨主机操作。
+
+* /another_rootfs/usr/local/bin/containerd-shim
+
+``` shell
+#!/bin/bash
+CMD_NET_ADDR=tcp://<服务端ip>:<rexec端口号> /usr/bin/rexec /usr/bin/containerd-shim $*
+```
+
+* /another_rootfs/usr/local/bin/remote_kill
+
+``` shell
+#!/bin/bash
+CMD_NET_ADDR=tcp://<服务端ip>:<rexec端口号> /usr/bin/rexec /usr/bin/kill $*
+```
+
+* /another_rootfs/usr/sbin/modprobe
+``` shell
+#!/bin/bash
+CMD_NET_ADDR=tcp://<服务端ip>:<rexec端口号> /usr/bin/rexec /usr/sbin/modprobe $*
+```
+
+在chroot到dockerd和containerd运行所需的rootfs后，用如下的命令拉起dockerd和containerd
+
+* containerd
+``` shell
+#!/bin/bash
+SHIM_HOST=<服务端ip> containerd --config /var/run/docker/containerd/containerd.toml --address /var/run/containerd/containerd.sock
+```
+
+* dockerd
+``` shell
+#!/bin/bash
+SHIM_HOST=<服务端ip> CMD_NET_ADDR=tcp://<服务端ip>:<rexec端口号> /usr/bin/dockerd --containerd /var/run/containerd/containerd.sock
+```
+
+* kubelet
+
+在chroot环境内使用原参数拉起kubelet即可。
+
+因为我们已经将/var/run/和/another_rootfs/var/run/绑定在了一起，所以可以在正常的rootfs下，通过docker来访问docker.sock接口进行容器管理。
+
+至此，完成容器管理面卸载到DPU，可以通过docker相关操作进行容器创建、删除等操作，也可以通过kubectl在当前节点进行pods调度和销毁，且实际容器业务进程运行在HOST侧。
+
+> ![](./public_sys-resources/icon-note.gif)**说明**：
+>
+> 本指导所述操作只涉及容器管理面进程卸载，不包含容器网络和数据卷volume等卸载，如有相关需求，需要通过额外的网络或存储卸载能力支持。本指导支持不带网络和存储的容器跨节点拉起。
\ No newline at end of file
--
2.33.0

56
0010-update-libvirt-doc.patch Normal file
View File

@ -0,0 +1,56 @@
From ab96a3f0279f6befda7c66ea80956defe4b90f40 Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Fri, 9 Jun 2023 15:20:31 +0800
Subject: update libvirt doc
Signed-off-by: liqiang <liqiang64@huawei.com>
---
...\351\233\266\346\220\255\345\273\272v1.1.md" | 17 ++++++++++++++---
1 file changed, 14 insertions(+), 3 deletions(-)
diff --git "a/usecases/libvirtd-offload/libvirt\347\233\264\350\277\236\350\201\232\345\220\210\347\216\257\345\242\203\344\273\216\351\233\266\346\220\255\345\273\272v1.1.md" "b/usecases/libvirtd-offload/libvirt\347\233\264\350\277\236\350\201\232\345\220\210\347\216\257\345\242\203\344\273\216\351\233\266\346\220\255\345\273\272v1.1.md"
index c1f9c87..85541dc 100644
--- "a/usecases/libvirtd-offload/libvirt\347\233\264\350\277\236\350\201\232\345\220\210\347\216\257\345\242\203\344\273\216\351\233\266\346\220\255\345\273\272v1.1.md"
+++ "b/usecases/libvirtd-offload/libvirt\347\233\264\350\277\236\350\201\232\345\220\210\347\216\257\345\242\203\344\273\216\351\233\266\346\220\255\345\273\272v1.1.md"
@@ -2,16 +2,27 @@
| ---- | --------- | ---- | ------------------------------------------------------------ |
| V1.0 | 2022/12/5 | 李强 | 创建文档 |
| V1.1 | 2023/2/6 | 李强 | 增加uds proxy组件部署介绍；增加rexec组件部署介绍；修改libvirt相关描述，适配uds proxy组件。 |
-| V1.2 | 2023/3/22 | 李强 | rexec重构后，更新rexec组件部署说明。 |
-| V1.3 | 2023/3/27 | 李强 | qtfs为connect syscall适配uds做改动 |
+| V1.2 | 2023/3/22 | 李强 | rexec重构后，更新rexec组件部署说明。 |
+| V1.3 | 2023/3/27 | 李强 | qtfs为connect syscall适配uds做改动。 |
+| V1.4 | 2023/6/9 | 李强 | 将默认的组件链接方式改为vsock，网络改为测试模式。 |
# **1** 硬件准备
+## 测试模式
+
需准备2台物理机（虚机当前未试过），网络互通。
其中一台作为DPU模拟，另一台作为HOST模拟。在本文档中用DPU和HOST指代这两台服务器。
+注意：测试模式因为会暴露网络端口且不做连接认证，存在网络安全风险，仅能用于内部测试验证，不要用于实际生产环境。
+
+## vsock模式
+
+需要DPU加HOST，且DPU能支持通过virtio提供vsock通信方式。
+
+目前还未基于真实的支持DPU vsock的环境调试过，本文档当前仅描述基于测试模式的方法，下面的内容依然默认使用测试模式。
+
# **2** libvirt卸载架构图
@@ -40,7 +51,7 @@ udsproxyd是一个跨主机的unix domain socket代理服务，需要分别部
```bash
cd qtfs/ipc
-make && make install
+make -j UDS_TEST_MODE=1 && make install
```
当前最新版本下，qtfs server侧的engine服务已经整合了udsproxyd的能力，所以server侧若部署了qtfs后不需要再额外启动udsproxyd。client侧则单独拉起udsproxyd服务：
--
2.33.0

286
0011-ioctl-enhance-and-rewrite-duplicate-code.patch Normal file
View File

@ -0,0 +1,286 @@
From bf30c2c3dc20b2b5e9562313a4c50541ebe313de Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Sat, 10 Jun 2023 11:30:37 +0800
Subject: ioctl enhance and rewrite duplicate code
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/include/req.h | 2 +
qtfs/qtfs/sb.c | 43 +++++++++-----
qtfs/qtfs_server/fsops.c | 122 ++++++++++++++++-----------------------
3 files changed, 81 insertions(+), 86 deletions(-)
diff --git a/qtfs/include/req.h b/qtfs/include/req.h
index 7e0a4b2..407faef 100644
--- a/qtfs/include/req.h
+++ b/qtfs/include/req.h
@@ -135,6 +135,8 @@ struct qtreq_ioctl {
unsigned int cmd;
unsigned int size;
int fd;
+ int argtype; // 0--use pointer arg, 1--use long arg
+ unsigned long arg; // for long type arg
} d;
char path[QTFS_TAIL_LEN(struct qtreq_ioctl_len)];
diff --git a/qtfs/qtfs/sb.c b/qtfs/qtfs/sb.c
index 2038c55..9c03219 100644
--- a/qtfs/qtfs/sb.c
+++ b/qtfs/qtfs/sb.c
@@ -20,7 +20,7 @@
#include <linux/version.h>
#include <asm-generic/ioctls.h>
#include <asm-generic/termbits.h>
-
+#include <linux/if_tun.h>
#include "conn.h"
#include "qtfs-mod.h"
@@ -561,7 +561,7 @@ int qtfs_fsync(struct file *file, loff_t start, loff_t end, int datasync)
return 0;
}
-long qtfs_do_ioctl(struct file *filp, unsigned int cmd, unsigned long arg, unsigned int size)
+long qtfs_do_ioctl(struct file *filp, unsigned int cmd, unsigned long arg, unsigned int size, int argtype)
{
struct qtfs_conn_var_s *pvar = qtfs_conn_get_param();
struct qtreq_ioctl *req;
@@ -575,19 +575,22 @@ long qtfs_do_ioctl(struct file *filp, unsigned int cmd, unsigned long arg, unsig
return -EINVAL;
}
- if (size >= MAX_PATH_LEN) {
- WARN_ON(1);
+ req = pvar->conn_ops->get_conn_msg_buf(pvar, QTFS_SEND);
+ rsp = pvar->conn_ops->get_conn_msg_buf(pvar, QTFS_RECV);
+ if (size >= sizeof(req->path)) {
+ qtfs_err("do ioctl failed, size:%u too big:%u", size, sizeof(req->path));
qtfs_conn_put_param(pvar);
return -EINVAL;
}
- req = pvar->conn_ops->get_conn_msg_buf(pvar, QTFS_SEND);
- rsp = pvar->conn_ops->get_conn_msg_buf(pvar, QTFS_RECV);
priv = (struct private_data *)filp->private_data;
req->d.fd = priv->fd;
-
+ req->d.argtype = argtype;
req->d.cmd = cmd;
- if (size > 0) {
+ if (argtype) {
+ req->d.arg = arg;
+ len = sizeof(struct qtreq_ioctl) - sizeof(req->path);
+ } else if (size > 0) {
ret = copy_from_user(req->path, (char __user *)arg, size);
if (ret) {
qtfs_err("%s: copy_from_user, size %u failed.", __func__, size);
@@ -597,7 +600,7 @@ long qtfs_do_ioctl(struct file *filp, unsigned int cmd, unsigned long arg, unsig
len = sizeof(struct qtreq_ioctl) - sizeof(req->path) + size;
req->d.size = size;
} else {
- len = sizeof(struct qtreq_ioctl) - sizeof(req->path) + strlen(req->path) + 1;
+ len = sizeof(struct qtreq_ioctl) - sizeof(req->path);
}
rsp = qtfs_remote_run(pvar, QTFS_REQ_IOCTL, len);
@@ -623,27 +626,41 @@ out:
return (long)ret;
}
+#define QTFS_IOCTL_CASE_WITH_BREAK(size, argtype)\
+ {\
+ ret = qtfs_do_ioctl(filp, cmd, arg, size, argtype);\
+ break;\
+ }
long qtfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
{
+ unsigned int size;
+ long ret;
switch(cmd) {
+ // all case of size 0 type 0 enter here
case FS_IOC_FSGETXATTR:
case TCGETS:
- return qtfs_do_ioctl(filp, cmd, arg, 0);
+ QTFS_IOCTL_CASE_WITH_BREAK(0, 0);
+ // all case of size 0 type 1 enter here
+ case TUNSETPERSIST:
+ QTFS_IOCTL_CASE_WITH_BREAK(0, 1);
case FS_IOC_FSSETXATTR:
- return qtfs_do_ioctl(filp, cmd, arg, sizeof(struct fsxattr));
+ QTFS_IOCTL_CASE_WITH_BREAK(sizeof(struct fsxattr), 0);
case TCSETS:
- return qtfs_do_ioctl(filp, cmd, arg, sizeof(struct ktermios));
+ QTFS_IOCTL_CASE_WITH_BREAK(sizeof(struct ktermios), 0);
+ case TUNSETIFF:
+ QTFS_IOCTL_CASE_WITH_BREAK(sizeof(struct ifreq), 0);
default: {
char *fullname = kmalloc(MAX_PATH_LEN, GFP_KERNEL);
if (!fullname)
return -ENOMEM;
memset(fullname, 0, MAX_PATH_LEN);
qtfs_fullname(fullname, filp->f_path.dentry, MAX_PATH_LEN);
- qtfs_err("qtfs ioctl get not support cmd:%d file:%s TCGETS:%d", cmd, fullname, TCGETS);
+ qtfs_err("qtfs ioctl get not support cmd:%d file:%s", cmd, fullname);
kfree(fullname);
return -EOPNOTSUPP;
}
}
+ return ret;
}
loff_t qtfs_dir_file_llseek(struct file *file, loff_t offset, int whence)
diff --git a/qtfs/qtfs_server/fsops.c b/qtfs/qtfs_server/fsops.c
index 0ae95f3..e8e5742 100644
--- a/qtfs/qtfs_server/fsops.c
+++ b/qtfs/qtfs_server/fsops.c
@@ -29,6 +29,7 @@
#include <linux/uio.h>
#include <linux/blkdev.h>
#include <linux/version.h>
+#include <linux/if_tun.h>
#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0))
#include <linux/fdtable.h>
#endif
@@ -105,8 +106,20 @@ static inline void qtfs_inode_info_fill(struct inode_info *ii, struct inode *ino
return;
}
+#define QTFS_IOCTL_HANDLE_WITH_BREAK(rspsize)\
+ {\
+ ret = copy_from_user(rsp->buf, userp->userp, rspsize);\
+ if (ret) {\
+ qtfs_err("cmd:%d copy_from_user failed with:%d\n", req->d.cmd, ret);\
+ rsp->errno = -EFAULT;\
+ goto err;\
+ }\
+ rsp->size = rspsize;\
+ break;\
+ }
static int handle_ioctl(struct qtserver_arg *arg)
{
+ unsigned long ioctl_arg;
int ret;
int iret;
struct qtreq_ioctl *req = (struct qtreq_ioctl *)REQ(arg);
@@ -123,82 +136,45 @@ static int handle_ioctl(struct qtserver_arg *arg)
}
mutex_unlock(&fd_bitmap_lock);
- switch (req->d.cmd) {
- case FS_IOC_FSGETXATTR:
- iret = qtfs_syscall_ioctl(req->d.fd, req->d.cmd, (unsigned long)userp->userp);
- if (iret) {
- qtfs_err("fsgetxattr ioctl failed with %d\n", iret);
- rsp->errno = iret;
- goto err;
- }
- ret = copy_from_user(rsp->buf, userp->userp, sizeof(struct fsxattr));
- if (ret) {
- qtfs_err("fsgetxattr copy_from_user failed with %d\n", ret);
- rsp->errno = -EFAULT;
- goto err;
- }
- rsp->size = sizeof(struct fsxattr);
- break;
- case FS_IOC_FSSETXATTR:
- if (req->d.size <= 0 || req->d.size > sizeof(req->path) || req->d.size >= userp->size) {
- rsp->errno = -EINVAL;
- goto err;
- }
- ret = copy_to_user(userp->userp, req->path, req->d.size);
- if (ret) {
- qtfs_err("fssetxattr copy_to_user failed with %d\n", ret);
- rsp->errno = -EFAULT;
- goto err;
- }
- iret = qtfs_syscall_ioctl(req->d.fd, req->d.cmd, (unsigned long)userp->userp);
- if (iret) {
- qtfs_err("fssetxattr ioctl failed with %d\n", iret);
- rsp->errno = iret;
- goto err;
- }
- rsp->size = 0;
- break;
- case TCGETS:
- iret = qtfs_syscall_ioctl(req->d.fd, req->d.cmd, (unsigned long)userp->userp);
- if (iret) {
- qtfs_err("ioctl TCGETS failed with %d\n", iret);
- rsp->errno = iret;
- goto err;
- }
- qtfs_info("ioctl TCGETS ret:%d", iret);
-
- ret = copy_from_user(rsp->buf, userp->userp, sizeof(struct ktermios));
- if (ret) {
- qtfs_err("fsgetxattr copy_from_user failed with %d\n", ret);
- rsp->errno = -EFAULT;
- goto err;
- }
- rsp->size = sizeof(struct ktermios);
- break;
- case TCSETS:
- if (req->d.size <= 0 || req->d.size > sizeof(req->path)) {
- rsp->errno = -EINVAL;
- goto err;
- }
- ret = copy_to_user(userp->userp, req->path, req->d.size);
- if (ret) {
- qtfs_err("tcsets copy_to_user failed with %d\n", ret);
- rsp->errno = -EFAULT;
- goto err;
- }
- qtfs_info("tcsets size:%u sizeof ktermios:%lu", req->d.size, sizeof(struct ktermios));
- iret = qtfs_syscall_ioctl(req->d.fd, req->d.cmd, (unsigned long)userp->userp);
- if (iret) {
- qtfs_err("tcsets ioctl failed with %d\n", iret);
- rsp->errno = iret;
- goto err;
+ if (req->d.argtype) {
+ ioctl_arg = req->d.arg;
+ } else {
+ if (req->d.size) {
+ if (req->d.size <= 0 || req->d.size > sizeof(req->path) || req->d.size >= userp->size) {
+ rsp->errno = -EINVAL;
+ goto err;
+ }
+ ret = copy_to_user(userp->userp, req->path, req->d.size);
+ if (ret) {
+ qtfs_err("cmd:%d copy_to_user failed with:%d", req->d.cmd, ret);
+ rsp->errno = -EFAULT;
+ goto err;
+ }
}
- rsp->size = 0;
- break;
- default:
- rsp->errno = -EOPNOTSUPP;
+ ioctl_arg = (unsigned long)userp->userp;
+ }
+ iret = qtfs_syscall_ioctl(req->d.fd, req->d.cmd, ioctl_arg);
+ if (iret) {
+ qtfs_err("ioctl fd:%d cmd:%d failed with %d", req->d.fd, req->d.cmd, iret);
+ rsp->errno = iret;
goto err;
}
+ qtfs_info("ioctl fd:%d cmd:%d argtype:%d arg:%lx size:%u successed", req->d.fd, req->d.cmd, req->d.argtype, req->d.arg, req->d.size);
+ switch (req->d.cmd) {
+ case TUNSETPERSIST:
+ case TUNSETIFF:
+ case TCSETS:
+ case FS_IOC_FSSETXATTR:
+ rsp->size = 0;
+ break;
+ case FS_IOC_FSGETXATTR:
+ QTFS_IOCTL_HANDLE_WITH_BREAK(sizeof(struct fsxattr));
+ case TCGETS:
+ QTFS_IOCTL_HANDLE_WITH_BREAK(sizeof(struct ktermios));
+ default:
+ rsp->errno = -EOPNOTSUPP;
+ goto err;
+ }
rsp->ret = QTFS_OK;
rsp->errno = iret;
return sizeof(struct qtrsp_ioctl) - sizeof(rsp->buf) + rsp->size;
--
2.33.0

127
0012-port-in-use.patch Normal file
View File

@ -0,0 +1,127 @@
From 890082069c4e43fd8d4ec6e96bf851d185195b7d Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Sat, 10 Jun 2023 17:02:53 +0800
Subject: port in use
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/include/conn.h | 12 ++++++++++--
qtfs/qtfs_common/conn.c | 10 +++++-----
qtfs/qtfs_common/user_engine.c | 5 ++++-
qtfs/qtfs_server/qtfs-server.c | 8 ++++++++
4 files changed, 27 insertions(+), 8 deletions(-)
diff --git a/qtfs/include/conn.h b/qtfs/include/conn.h
index 3b8f1f4..afbb00f 100644
--- a/qtfs/include/conn.h
+++ b/qtfs/include/conn.h
@@ -46,7 +46,16 @@ extern bool qtfs_epoll_mode;
extern struct qtsock_wl_stru qtsock_wl;
extern struct qtfs_pvar_ops_s qtfs_conn_sock_pvar_ops;
-#define qtfs_conn_get_param(void) _qtfs_conn_get_param(__func__)
+struct qtfs_conn_var_s *_qtfs_conn_get_param(const char *);
+static inline struct qtfs_conn_var_s *__qtfs_conn_get_param(const char *who_using)
+{
+ struct qtfs_conn_var_s *p = _qtfs_conn_get_param(who_using);
+ if (IS_ERR_OR_NULL(p))
+ return NULL;
+ return p;
+}
+#define qtfs_conn_get_param(void) __qtfs_conn_get_param(__func__)
+#define qtfs_conn_get_param_errcode(void) _qtfs_conn_get_param(__func__)
#define QTFS_CONN_SOCK_TYPE "socket"
#define QTFS_CONN_PCIE_TYPE "pcie"
@@ -172,7 +181,6 @@ void *qtfs_conn_msg_buf(struct qtfs_conn_var_s *pvar, int dir);
void qtfs_conn_param_init(void);
void qtfs_conn_param_fini(void);
-struct qtfs_conn_var_s *_qtfs_conn_get_param(const char *);
void qtfs_conn_put_param(struct qtfs_conn_var_s *pvar);
struct qtfs_conn_var_s *qtfs_epoll_establish_conn(void);
void qtfs_epoll_cut_conn(struct qtfs_conn_var_s *pvar);
diff --git a/qtfs/qtfs_common/conn.c b/qtfs/qtfs_common/conn.c
index 56b2b12..ab84181 100644
--- a/qtfs/qtfs_common/conn.c
+++ b/qtfs/qtfs_common/conn.c
@@ -580,7 +580,7 @@ retry:
}
if (ret != 0) {
qtfs_conn_put_param(pvar);
- return NULL;
+ return (IS_ERR_VALUE((long)ret) ? ERR_PTR((long)ret) : NULL);
}
strlcpy(pvar->who_using, func, QTFS_FUNCTION_LEN);
return pvar;
@@ -636,18 +636,18 @@ retry:
qtfs_err("qtfs get param active connection failed, ret:%d, curstate:%s", ret, QTCONN_CUR_STATE(pvar));
// put to vld list
qtfs_conn_put_param(pvar);
- return NULL;
+ return (IS_ERR_VALUE((long)ret) ? ERR_PTR((long)ret) : NULL);
}
qtfs_thread_var[pvar->cur_threadidx] = pvar;
#else
pvar->cs = QTFS_CONN_SOCK_SERVER;
if (!pvar->conn_ops->conn_inited(pvar)) {
- if (qtfs_sm_active(pvar)) {
+ if ((ret = qtfs_sm_active(pvar)) != 0) {
qtfs_err("qtfs get param active connection failed, ret:%d, curstate:%s", ret, QTCONN_CUR_STATE(pvar));
// put to vld list
mutex_unlock(&g_param_mutex);
qtfs_conn_put_param(pvar);
- return NULL;
+ return (IS_ERR_VALUE((long)ret) ? ERR_PTR((long)ret) : NULL);
}
mutex_unlock(&g_param_mutex);
} else {
@@ -657,7 +657,7 @@ retry:
if (ret) {
qtfs_err("qtfs get param active connection failed, ret:%d curstate:%s", ret, QTCONN_CUR_STATE(pvar));
qtfs_conn_put_param(pvar);
- return NULL;
+ return (IS_ERR_VALUE((long)ret) ? ERR_PTR((long)ret) : NULL);
}
}
#endif
diff --git a/qtfs/qtfs_common/user_engine.c b/qtfs/qtfs_common/user_engine.c
index 521ebe4..919da11 100644
--- a/qtfs/qtfs_common/user_engine.c
+++ b/qtfs/qtfs_common/user_engine.c
@@ -105,9 +105,12 @@ static struct qtfs_server_userp_s *qtfs_engine_thread_init(int fd, int thread_nu
}
}
struct qtfs_thread_init_s init_userp;
+ int ret;
init_userp.thread_nums = thread_nums;
init_userp.userp = userp;
- if (ioctl(fd, QTFS_IOCTL_THREAD_INIT, (unsigned long)&init_userp) == QTERROR) {
+ ret = ioctl(fd, QTFS_IOCTL_THREAD_INIT, (unsigned long)&init_userp);
+ if (ret != QTOK) {
+ engine_err("Engine thread init failed reason:%s", (ret == EADDRINUSE) ? strerror(EADDRINUSE) : "userp init failed.");
goto rollback;
}
return userp;
diff --git a/qtfs/qtfs_server/qtfs-server.c b/qtfs/qtfs_server/qtfs-server.c
index 003c361..729097b 100644
--- a/qtfs/qtfs_server/qtfs-server.c
+++ b/qtfs/qtfs_server/qtfs-server.c
@@ -192,6 +192,14 @@ long qtfs_server_misc_ioctl(struct file *file, unsigned int cmd, unsigned long a
struct qtfs_thread_init_s init_userp;
switch (cmd) {
case QTFS_IOCTL_THREAD_INIT:
+ pvar = qtfs_conn_get_param_errcode();
+ if (IS_ERR_OR_NULL(pvar)) {
+ qtfs_err("init pvar get failed, pvar:%ld", (long)pvar);
+ if (PTR_ERR(pvar) == -EADDRINUSE)
+ return EADDRINUSE;
+ } else {
+ qtfs_conn_put_param(pvar);
+ }
if (!write_trylock(&g_userp_rwlock)) {
qtfs_err("try lock userps failed.");
return QTERROR;
--
2.33.0

39
0013-ioctl-add-cmd.patch Normal file
View File

@ -0,0 +1,39 @@
From 31262c9f982d59bf7f407f2f698933515dc4c8c7 Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Mon, 12 Jun 2023 11:28:27 +0800
Subject: ioctl add cmd
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/qtfs/sb.c | 1 +
qtfs/qtfs_server/fsops.c | 2 ++
2 files changed, 3 insertions(+)
diff --git a/qtfs/qtfs/sb.c b/qtfs/qtfs/sb.c
index 9c03219..f557e6d 100644
--- a/qtfs/qtfs/sb.c
+++ b/qtfs/qtfs/sb.c
@@ -648,6 +648,7 @@ long qtfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
case TCSETS:
QTFS_IOCTL_CASE_WITH_BREAK(sizeof(struct ktermios), 0);
case TUNSETIFF:
+ case SIOCGIFHWADDR:
QTFS_IOCTL_CASE_WITH_BREAK(sizeof(struct ifreq), 0);
default: {
char *fullname = kmalloc(MAX_PATH_LEN, GFP_KERNEL);
diff --git a/qtfs/qtfs_server/fsops.c b/qtfs/qtfs_server/fsops.c
index e8e5742..727c4e1 100644
--- a/qtfs/qtfs_server/fsops.c
+++ b/qtfs/qtfs_server/fsops.c
@@ -171,6 +171,8 @@ static int handle_ioctl(struct qtserver_arg *arg)
QTFS_IOCTL_HANDLE_WITH_BREAK(sizeof(struct fsxattr));
case TCGETS:
QTFS_IOCTL_HANDLE_WITH_BREAK(sizeof(struct ktermios));
+ case SIOCGIFHWADDR:
+ QTFS_IOCTL_HANDLE_WITH_BREAK(sizeof(struct ifreq));
default:
rsp->errno = -EOPNOTSUPP;
goto err;
--
2.33.0

54
0014-engine-running-lock.patch Normal file
View File

@ -0,0 +1,54 @@
From a79ba9896c37d30cd1901f7f54e0507adb9aabda Mon Sep 17 00:00:00 2001
From: liqiang <liqiang64@huawei.com>
Date: Mon, 12 Jun 2023 11:34:09 +0800
Subject: engine running lock
Signed-off-by: liqiang <liqiang64@huawei.com>
---
qtfs/qtfs_common/user_engine.c | 16 +++++++++++++++-
1 file changed, 15 insertions(+), 1 deletion(-)
diff --git a/qtfs/qtfs_common/user_engine.c b/qtfs/qtfs_common/user_engine.c
index 919da11..7455413 100644
--- a/qtfs/qtfs_common/user_engine.c
+++ b/qtfs/qtfs_common/user_engine.c
@@ -33,7 +33,7 @@
#include <errno.h>
#include <sys/resource.h>
#include <sys/prctl.h>
-
+#include <sys/file.h>
#include <sys/epoll.h>
#include "comm.h"
@@ -68,6 +68,16 @@ struct engine_arg {
#define QTFS_USERP_SIZE QTFS_USERP_MAXSIZE
#define QTFS_SERVER_FILE "/dev/qtfs_server"
+#define ENGINE_LOCK_ADDR "/var/run/qtfs/engine.lock"
+
+int engine_socket_lock(void)
+{
+ int lock_fd = open(ENGINE_LOCK_ADDR, O_RDONLY | O_CREAT, 0600);
+ if (lock_fd == -1)
+ return -EINVAL;
+
+ return flock(lock_fd, LOCK_EX | LOCK_NB);
+}
int qtfs_fd;
int engine_run = 1;
@@ -365,6 +375,10 @@ int main(int argc, char *argv[])
engine_out(" Example: %s 16 1 192.168.10.10 12121 192.168.10.11 12121.", argv[0]);
return -1;
}
+ if (engine_socket_lock() < 0) {
+ engine_err("Engine is running.");
+ return -1;
+ }
if (qtfs_engine_env_check(argv) < 0) {
engine_err("Environment check failed, engine exit.");
return -1;
--
2.33.0

20
dpu-utilities.spec
View File

@ -1,7 +1,7 @@
Name: dpu-utilities
Summary: openEuler dpu utilities
Version: 1.4
Release: 1
Release: 2
License: GPL-2.0
Source: https://gitee.com/openeuler/dpu-utilities/repository/archive/v%{version}.tar.gz
ExclusiveOS: linux
@ -12,6 +12,21 @@ Provides: %{name} = %{version}-%{release}
%define kernel_version %(ver=`rpm -qa|grep kernel-devel`;echo ${ver#*kernel-devel-})
BuildRequires: kernel-devel >= 5.10, gcc, make, json-c-devel, glib2-devel
Patch1: 0001-rewrite-client-rexec_run.patch
Patch2: 0002-uds-add-vsock-mode.patch
Patch3: 0003-fix-a-rexec_server-bug-on-child-process.patch
Patch4: 0004-add-uds-test-mode-in-engine-compile.patch
Patch5: 0005-modify-qtfs_dentry_revalidate.patch
Patch6: 0006-add-cmake-compile-for-user-binary-and-so.patch
Patch7: 0007-uds-remove-MSG_WAITALL-flags-to-timeout.patch
Patch8: 0008-Modify-the-logic-of-rexec-proxy-process-exit-by-mist.patch
Patch9: 0009-Keep-doc-same-with-docs-in-openEuler.patch
Patch10: 0010-update-libvirt-doc.patch
Patch11: 0011-ioctl-enhance-and-rewrite-duplicate-code.patch
Patch12: 0012-port-in-use.patch
Patch13: 0013-ioctl-add-cmd.patch
Patch14: 0014-engine-running-lock.patch
%description
This package contains the software utilities on dpu.
@ -131,6 +146,9 @@ sed -i '/# product cut_conf/a\dpuos kiwi/minios/cfg_dpuos yes' /opt/imageT
sed -i '/<repository_rule>/a\dpuos 1 rpm-dir euler_base' /opt/imageTailor/repos/RepositoryRule.conf
%changelog
* Mon Jun 12 2023 Weifeng Su <suweifeng1@huawei.com> 1.4-2
- Sync patches from source
* Fri Jun 2 2023 Weifeng Su <suweifeng1@huawei.com> 1.4-1
- Upgrade dpu-utilities version to 1.4