Diff below is the first of 3 steps to unlock most of the network
syscalls. It deals with the simpler problem of protecting statistic
fields of the descriptor. For that I'm using a simple IPL_NONE mutex.
Diff below also documents how the various fields are protected. In the
next steps I'll deal with the reference counts, mainly `f_count' and the
one in `filehead' then the associated `f_data' pointer.
Note that the diff below doesn't deal with `f_offset' because it is more
tricky and not needed for recvmsg(2), sendmsg(2) & friends which are my
targets at the moment.
Comments? Ok?
Index: kern/kern_descrip.c
===================================================================
RCS file: /cvs/src/sys/kern/kern_descrip.c,v
retrieving revision 1.156
diff -u -p -r1.156 kern_descrip.c
--- kern/kern_descrip.c 2 May 2018 02:24:56 -0000 1.156
+++ kern/kern_descrip.c 7 May 2018 16:47:58 -0000
@@ -957,6 +957,7 @@ restart:
*/
numfiles++;
fp = pool_get(&file_pool, PR_WAITOK|PR_ZERO);
+ mtx_init(&fp->f_mtx, IPL_NONE);
fp->f_iflags = FIF_LARVAL;
if ((fq = p->p_fd->fd_ofiles[0]) != NULL) {
LIST_INSERT_AFTER(fq, fp, f_list);
Index: kern/kern_sysctl.c
===================================================================
RCS file: /cvs/src/sys/kern/kern_sysctl.c,v
retrieving revision 1.333
diff -u -p -r1.333 kern_sysctl.c
--- kern/kern_sysctl.c 25 Apr 2018 10:29:16 -0000 1.333
+++ kern/kern_sysctl.c 7 May 2018 16:15:18 -0000
@@ -1067,11 +1067,13 @@ fill_file(struct kinfo_file *kf, struct
if (suser(p) == 0 || p->p_ucred->cr_uid == fp->f_cred->cr_uid) {
kf->f_offset = fp->f_offset;
+ mtx_enter(&fp->f_mtx);
kf->f_rxfer = fp->f_rxfer;
kf->f_rwfer = fp->f_wxfer;
kf->f_seek = fp->f_seek;
kf->f_rbytes = fp->f_rbytes;
kf->f_wbytes = fp->f_wbytes;
+ mtx_leave(&fp->f_mtx);
} else
kf->f_offset = -1;
} else if (vp != NULL) {
Index: kern/sys_generic.c
===================================================================
RCS file: /cvs/src/sys/kern/sys_generic.c,v
retrieving revision 1.118
diff -u -p -r1.118 sys_generic.c
--- kern/sys_generic.c 27 Apr 2018 10:13:37 -0000 1.118
+++ kern/sys_generic.c 7 May 2018 16:13:02 -0000
@@ -203,8 +203,10 @@ dofilereadv(struct proc *p, int fd, stru
error = 0;
cnt -= auio.uio_resid;
+ mtx_enter(&fp->f_mtx);
fp->f_rxfer++;
fp->f_rbytes += cnt;
+ mtx_leave(&fp->f_mtx);
#ifdef KTRACE
if (ktriov != NULL) {
if (error == 0)
@@ -355,8 +357,10 @@ dofilewritev(struct proc *p, int fd, str
}
cnt -= auio.uio_resid;
+ mtx_enter(&fp->f_mtx);
fp->f_wxfer++;
fp->f_wbytes += cnt;
+ mtx_leave(&fp->f_mtx);
#ifdef KTRACE
if (ktriov != NULL) {
if (error == 0)
Index: kern/uipc_syscalls.c
===================================================================
RCS file: /cvs/src/sys/kern/uipc_syscalls.c,v
retrieving revision 1.169
diff -u -p -r1.169 uipc_syscalls.c
--- kern/uipc_syscalls.c 27 Apr 2018 10:13:37 -0000 1.169
+++ kern/uipc_syscalls.c 7 May 2018 16:13:40 -0000
@@ -687,8 +687,10 @@ sendit(struct proc *p, int s, struct msg
}
if (error == 0) {
*retsize = len - auio.uio_resid;
+ mtx_enter(&fp->f_mtx);
fp->f_wxfer++;
fp->f_wbytes += *retsize;
+ mtx_leave(&fp->f_mtx);
}
#ifdef KTRACE
if (ktriov != NULL) {
@@ -902,8 +904,10 @@ recvit(struct proc *p, int s, struct msg
mp->msg_controllen = len;
}
if (!error) {
+ mtx_enter(&fp->f_mtx);
fp->f_rxfer++;
fp->f_rbytes += *retsize;
+ mtx_leave(&fp->f_mtx);
}
out:
FRELE(fp, p);
Index: kern/vfs_syscalls.c
===================================================================
RCS file: /cvs/src/sys/kern/vfs_syscalls.c,v
retrieving revision 1.282
diff -u -p -r1.282 vfs_syscalls.c
--- kern/vfs_syscalls.c 2 May 2018 02:24:56 -0000 1.282
+++ kern/vfs_syscalls.c 7 May 2018 16:13:02 -0000
@@ -1650,7 +1650,9 @@ sys_lseek(struct proc *p, void *v, regis
}
}
*(off_t *)retval = fp->f_offset = newoff;
+ mtx_enter(&fp->f_mtx);
fp->f_seek++;
+ mtx_leave(&fp->f_mtx);
error = 0;
bad:
FRELE(fp, p);
Index: sys/file.h
===================================================================
RCS file: /cvs/src/sys/sys/file.h,v
retrieving revision 1.41
diff -u -p -r1.41 file.h
--- sys/file.h 25 Apr 2018 10:29:16 -0000 1.41
+++ sys/file.h 7 May 2018 16:52:11 -0000
@@ -37,6 +37,7 @@
#else /* _KERNEL */
#include <sys/queue.h>
+#include <sys/mutex.h>
struct proc;
struct uio;
@@ -61,26 +62,32 @@ struct fileops {
/*
* Kernel descriptor table.
* One entry for each open kernel vnode and socket.
+ *
+ * Locks used to protect struct members in this file:
+ * I immutable after creation
+ * f per file `f_mtx'
+ * k kernel lock
*/
struct file {
- LIST_ENTRY(file) f_list;/* list of active files */
- short f_flag; /* see fcntl.h */
+ LIST_ENTRY(file) f_list;/* [k] list of active files */
+ struct mutex f_mtx;
+ short f_flag; /* [I] see fcntl.h */
#define DTYPE_VNODE 1 /* file */
#define DTYPE_SOCKET 2 /* communications endpoint */
#define DTYPE_PIPE 3 /* pipe */
#define DTYPE_KQUEUE 4 /* event queue */
- short f_type; /* descriptor type */
- long f_count; /* reference count */
- struct ucred *f_cred; /* credentials associated with descriptor */
- struct fileops *f_ops;
- off_t f_offset;
- void *f_data; /* private data */
- int f_iflags; /* internal flags */
- u_int64_t f_rxfer; /* total number of read transfers */
- u_int64_t f_wxfer; /* total number of write transfers */
- u_int64_t f_seek; /* total independent seek operations */
- u_int64_t f_rbytes; /* total bytes read */
- u_int64_t f_wbytes; /* total bytes written */
+ short f_type; /* [I] descriptor type */
+ long f_count; /* [k] reference count */
+ struct ucred *f_cred; /* [I] credentials associated with descriptor */
+ struct fileops *f_ops; /* [I] file operation pointers */
+ off_t f_offset; /* [k] */
+ void *f_data; /* [k] private data */
+ int f_iflags; /* [k] internal flags */
+ uint64_t f_rxfer; /* [f] total number of read transfers */
+ uint64_t f_wxfer; /* [f] total number of write transfers */
+ uint64_t f_seek; /* [f] total independent seek operations */
+ uint64_t f_rbytes; /* [f] total bytes read */
+ uint64_t f_wbytes; /* [f] total bytes written */
};
#define FIF_HASLOCK 0x01 /* descriptor holds advisory lock */
