[CRIU] Re: [PATCH RFC] IPC: message queue stealing feature introduced

[Pavel Emelyanov]

On 02/10/2012 09:33 PM, Stanislav Kinsbursky wrote:
> This patch is required for checkpoint/restart ins userspace.
> IOW, c/r reuqires some way to get IPC messages without deleting them for the
> queue (checkpoint can fail and in this case tasks will be resumed, so queue
> have to be valid).
> To achive this, new operation flag MSG_STEAL for sys_msgrcv() system call
> introduced.
> If this flag is set, then passed struct msgbuf pointer will be used for storing
> array of structures:
> 
> struct msgbuf_a {
> 	long mtype;         /* type of message */
> 	size_t msize;       /* size of message */
> 	char mtext[0];      /* message text */
> };
> 
> where each of them is followed by respective message array.
> 
> Signed-off-by: Stanislav Kinsbursky <skinsbursky at parallels.com>

Ack. Cyrill, plz, put to our tree.

> ---
>  include/linux/msg.h |    8 ++++++++
>  ipc/msg.c           |   43 ++++++++++++++++++++++++++++++++++++++-----
>  2 files changed, 46 insertions(+), 5 deletions(-)
> 
> diff --git a/include/linux/msg.h b/include/linux/msg.h
> index 6689e73..2bace1a 100644
> --- a/include/linux/msg.h
> +++ b/include/linux/msg.h
> @@ -11,6 +11,7 @@
>  /* msgrcv options */
>  #define MSG_NOERROR     010000  /* no error if message is too big */
>  #define MSG_EXCEPT      020000  /* recv any msg except of specified type.*/
> +#define MSG_STEAL       040000  /* copy (not remove) all queue messages */
>  
>  /* Obsolete, used only for backwards compatibility and libc5 compiles */
>  struct msqid_ds {
> @@ -38,6 +39,13 @@ struct msgbuf {
>  	char mtext[1];      /* message text */
>  };
>  
> +/* message buffer for msgrcv in case of array calls */
> +struct msgbuf_a {
> +	long mtype;         /* type of message */
> +	size_t msize;       /* size of message */
> +	char mtext[0];      /* message text */
> +};
> +
>  /* buffer for msgctl calls IPC_INFO, MSG_INFO */
>  struct msginfo {
>  	int msgpool;
> diff --git a/ipc/msg.c b/ipc/msg.c
> index 303362b..d024cdc 100644
> --- a/ipc/msg.c
> +++ b/ipc/msg.c
> @@ -762,6 +762,23 @@ static inline int convert_mode(long *msgtyp, int msgflg)
>  	return SEARCH_EQUAL;
>  }
>  
> +static int do_store_msg(void __user *mtext, size_t msgsz, struct msg_msg *msg)
> +{
> +	msgsz = (msgsz > msg->m_ts) ? msg->m_ts : msgsz;
> +	if (store_msg(mtext, msg, msgsz))
> +		return -EFAULT;
> +	return msgsz;
> +}
> +
> +static int do_steal_msg(struct msgbuf_a __user *msgp, size_t msgsz, struct msg_msg *msg)
> +{
> +	if (put_user(msg->m_type, &msgp->mtype))
> +		return -EFAULT;
> +	if (put_user(msg->m_ts, &msgp->msize))
> +		return -EFAULT;
> +	return do_store_msg(msgp->mtext, msgsz, msg);
> +}
> +
>  long do_msgrcv(int msqid, long *pmtype, void __user *mtext,
>  		size_t msgsz, long msgtyp, int msgflg)
>  {
> @@ -769,6 +786,7 @@ long do_msgrcv(int msqid, long *pmtype, void __user *mtext,
>  	struct msg_msg *msg;
>  	int mode;
>  	struct ipc_namespace *ns;
> +	size_t arrsz = msgsz;
>  
>  	if (msqid < 0 || (long) msgsz < 0)
>  		return -EINVAL;
> @@ -802,6 +820,16 @@ long do_msgrcv(int msqid, long *pmtype, void __user *mtext,
>  						walk_msg->m_type != 1) {
>  					msg = walk_msg;
>  					msgtyp = walk_msg->m_type - 1;
> +				} else if (msgflg & MSG_STEAL) {
> +					int ret;
> +
> +					ret = do_steal_msg(mtext, arrsz, msg);
> +					if (ret < 0) {
> +						msg = ERR_PTR(ret);
> +						goto out_unlock;
> +					}
> +					mtext += ret + sizeof(struct msgbuf_a);
> +					arrsz -= ret + sizeof(struct msgbuf_a);
>  				} else {
>  					msg = walk_msg;
>  					break;
> @@ -810,6 +838,8 @@ long do_msgrcv(int msqid, long *pmtype, void __user *mtext,
>  			tmp = tmp->next;
>  		}
>  		if (!IS_ERR(msg)) {
> +			if (msgflg & MSG_STEAL)
> +				goto out_unlock;
>  			/*
>  			 * Found a suitable message.
>  			 * Unlink it from the queue.
> @@ -904,10 +934,11 @@ out_unlock:
>  	if (IS_ERR(msg))
>  		return PTR_ERR(msg);
>  
> -	msgsz = (msgsz > msg->m_ts) ? msg->m_ts : msgsz;
> +	if (msgflg & MSG_STEAL)
> +		return msgsz;
> +
>  	*pmtype = msg->m_type;
> -	if (store_msg(mtext, msg, msgsz))
> -		msgsz = -EFAULT;
> +	msgsz = do_store_msg(mtext, msgsz, msg);
>  
>  	free_msg(msg);
>  
> @@ -919,8 +950,10 @@ SYSCALL_DEFINE5(msgrcv, int, msqid, struct msgbuf __user *, msgp, size_t, msgsz,
>  {
>  	long err, mtype;
>  
> -	err =  do_msgrcv(msqid, &mtype, msgp->mtext, msgsz, msgtyp, msgflg);
> -	if (err < 0)
> +	err =  do_msgrcv(msqid, &mtype,
> +			 (msgflg & MSG_STEAL) ? (void *)msgp : msgp->mtext,
> +			 msgsz, msgtyp, msgflg);
> +	if (err < 0 || msgflg & MSG_STEAL)
>  		goto out;
>  
>  	if (put_user(mtype, &msgp->mtype))
>