[ltt-dev] [LTTNG-TOOLS PATCH 1/2] Function to handle snapshots in liblttngkconsumerd

[Mathieu Desnoyers]

* Julien Desfossez (julien.desfossez at polymtl.ca) wrote:
> This patchs adds the wrappers to take a snapshot for a specific FD and
> to retreive the produced position.
> Also refactor the sync tracefile operation.
> 

merged, thanks!

Mathieu

> Signed-off-by: Julien Desfossez <julien.desfossez at polymtl.ca>
> ---
>  include/lttng/lttng-kconsumerd.h     |   18 ++++
>  liblttngkconsumerd/lttngkconsumerd.c |  142 +++++++++++++++++++---------------
>  2 files changed, 98 insertions(+), 62 deletions(-)
> 
> diff --git a/include/lttng/lttng-kconsumerd.h b/include/lttng/lttng-kconsumerd.h
> index 8d16d56..7e195ab 100644
> --- a/include/lttng/lttng-kconsumerd.h
> +++ b/include/lttng/lttng-kconsumerd.h
> @@ -132,6 +132,24 @@ extern int lttng_kconsumerd_on_read_subbuffer_splice(
>  		struct lttng_kconsumerd_fd *kconsumerd_fd, unsigned long len);
>  
>  /*
> + * Take a snapshot for a specific fd
> + *
> + * Returns 0 on success, < 0 on error
> + */
> +int lttng_kconsumerd_take_snapshot(struct lttng_kconsumerd_local_data *ctx,
> +        struct lttng_kconsumerd_fd *kconsumerd_fd);
> +
> +/*
> + * Get the produced position
> + *
> + * Returns 0 on success, < 0 on error
> + */
> +int lttng_kconsumerd_get_produced_snapshot(
> +        struct lttng_kconsumerd_local_data *ctx,
> +        struct lttng_kconsumerd_fd *kconsumerd_fd,
> +        unsigned long *pos);
> +
> +/*
>   * Send return code to session daemon.
>   *
>   * Returns the return code of sendmsg : the number of bytes transmitted or -1
> diff --git a/liblttngkconsumerd/lttngkconsumerd.c b/liblttngkconsumerd/lttngkconsumerd.c
> index ba26026..d36da9b 100644
> --- a/liblttngkconsumerd/lttngkconsumerd.c
> +++ b/liblttngkconsumerd/lttngkconsumerd.c
> @@ -361,6 +361,42 @@ void lttng_kconsumerd_set_command_sock_path(
>  	ctx->kconsumerd_command_sock_path = sock;
>  }
>  
> +static void lttng_kconsumerd_sync_trace_file(
> +		struct lttng_kconsumerd_fd *kconsumerd_fd, off_t orig_offset)
> +{
> +	int outfd = kconsumerd_fd->out_fd;
> +	/*
> +	 * This does a blocking write-and-wait on any page that belongs to the
> +	 * subbuffer prior to the one we just wrote.
> +	 * Don't care about error values, as these are just hints and ways to
> +	 * limit the amount of page cache used.
> +	 */
> +	if (orig_offset >= kconsumerd_fd->max_sb_size) {
> +		sync_file_range(outfd, orig_offset - kconsumerd_fd->max_sb_size,
> +				kconsumerd_fd->max_sb_size,
> +				SYNC_FILE_RANGE_WAIT_BEFORE
> +				| SYNC_FILE_RANGE_WRITE
> +				| SYNC_FILE_RANGE_WAIT_AFTER);
> +		/*
> +		 * Give hints to the kernel about how we access the file:
> +		 * POSIX_FADV_DONTNEED : we won't re-access data in a near future after
> +		 * we write it.
> +		 *
> +		 * We need to call fadvise again after the file grows because the
> +		 * kernel does not seem to apply fadvise to non-existing parts of the
> +		 * file.
> +		 *
> +		 * Call fadvise _after_ having waited for the page writeback to
> +		 * complete because the dirty page writeback semantic is not well
> +		 * defined. So it can be expected to lead to lower throughput in
> +		 * streaming.
> +		 */
> +		posix_fadvise(outfd, orig_offset - kconsumerd_fd->max_sb_size,
> +				kconsumerd_fd->max_sb_size, POSIX_FADV_DONTNEED);
> +	}
> +}
> +
> +
>  /*
>   * Mmap the ring buffer, read it and write the data to the tracefile.
>   *
> @@ -371,7 +407,6 @@ int lttng_kconsumerd_on_read_subbuffer_mmap(
>  		struct lttng_kconsumerd_fd *kconsumerd_fd, unsigned long len)
>  {
>  	unsigned long mmap_offset;
> -	char *padding = NULL;
>  	long ret = 0;
>  	off_t orig_offset = kconsumerd_fd->out_fd_offset;
>  	int fd = kconsumerd_fd->consumerd_fd;
> @@ -400,42 +435,11 @@ int lttng_kconsumerd_on_read_subbuffer_mmap(
>  		kconsumerd_fd->out_fd_offset += ret;
>  	}
>  
> -	/*
> -	 * This does a blocking write-and-wait on any page that belongs to the
> -	 * subbuffer prior to the one we just wrote.
> -	 * Don't care about error values, as these are just hints and ways to
> -	 * limit the amount of page cache used.
> -	 */
> -	if (orig_offset >= kconsumerd_fd->max_sb_size) {
> -		sync_file_range(outfd, orig_offset - kconsumerd_fd->max_sb_size,
> -				kconsumerd_fd->max_sb_size,
> -				SYNC_FILE_RANGE_WAIT_BEFORE
> -				| SYNC_FILE_RANGE_WRITE
> -				| SYNC_FILE_RANGE_WAIT_AFTER);
> +	lttng_kconsumerd_sync_trace_file(kconsumerd_fd, orig_offset);
>  
> -		/*
> -		 * Give hints to the kernel about how we access the file:
> -		 * POSIX_FADV_DONTNEED : we won't re-access data in a near future after
> -		 * we write it.
> -		 *
> -		 * We need to call fadvise again after the file grows because the
> -		 * kernel does not seem to apply fadvise to non-existing parts of the
> -		 * file.
> -		 *
> -		 * Call fadvise _after_ having waited for the page writeback to
> -		 * complete because the dirty page writeback semantic is not well
> -		 * defined. So it can be expected to lead to lower throughput in
> -		 * streaming.
> -		 */
> -		posix_fadvise(outfd, orig_offset - kconsumerd_fd->max_sb_size,
> -				kconsumerd_fd->max_sb_size, POSIX_FADV_DONTNEED);
> -	}
>  	goto end;
>  
>  end:
> -	if (padding != NULL) {
> -		free(padding);
> -	}
>  	return ret;
>  }
>  
> @@ -482,36 +486,8 @@ int lttng_kconsumerd_on_read_subbuffer_splice(
>  				SYNC_FILE_RANGE_WRITE);
>  		kconsumerd_fd->out_fd_offset += ret;
>  	}
> +	lttng_kconsumerd_sync_trace_file(kconsumerd_fd, orig_offset);
>  
> -	/*
> -	 * This does a blocking write-and-wait on any page that belongs to the
> -	 * subbuffer prior to the one we just wrote.
> -	 * Don't care about error values, as these are just hints and ways to
> -	 * limit the amount of page cache used.
> -	 */
> -	if (orig_offset >= kconsumerd_fd->max_sb_size) {
> -		sync_file_range(outfd, orig_offset - kconsumerd_fd->max_sb_size,
> -				kconsumerd_fd->max_sb_size,
> -				SYNC_FILE_RANGE_WAIT_BEFORE
> -				| SYNC_FILE_RANGE_WRITE
> -				| SYNC_FILE_RANGE_WAIT_AFTER);
> -		/*
> -		 * Give hints to the kernel about how we access the file:
> -		 * POSIX_FADV_DONTNEED : we won't re-access data in a near future after
> -		 * we write it.
> -		 *
> -		 * We need to call fadvise again after the file grows because the
> -		 * kernel does not seem to apply fadvise to non-existing parts of the
> -		 * file.
> -		 *
> -		 * Call fadvise _after_ having waited for the page writeback to
> -		 * complete because the dirty page writeback semantic is not well
> -		 * defined. So it can be expected to lead to lower throughput in
> -		 * streaming.
> -		 */
> -		posix_fadvise(outfd, orig_offset - kconsumerd_fd->max_sb_size,
> -				kconsumerd_fd->max_sb_size, POSIX_FADV_DONTNEED);
> -	}
>  	goto end;
>  
>  splice_error:
> @@ -536,6 +512,48 @@ end:
>  }
>  
>  /*
> + * Take a snapshot for a specific fd
> + *
> + * Returns 0 on success, < 0 on error
> + */
> +int lttng_kconsumerd_take_snapshot(struct lttng_kconsumerd_local_data *ctx,
> +		struct lttng_kconsumerd_fd *kconsumerd_fd)
> +{
> +	int ret = 0;
> +	int infd = kconsumerd_fd->consumerd_fd;
> +
> +	ret = kernctl_snapshot(infd);
> +	if (ret != 0) {
> +		ret = errno;
> +		perror("Getting sub-buffer snapshot.");
> +	}
> +
> +	return ret;
> +}
> +
> +/*
> + * Get the produced position
> + *
> + * Returns 0 on success, < 0 on error
> + */
> +int lttng_kconsumerd_get_produced_snapshot(
> +		struct lttng_kconsumerd_local_data *ctx,
> +		struct lttng_kconsumerd_fd *kconsumerd_fd,
> +		unsigned long *pos)
> +{
> +	int ret;
> +	int infd = kconsumerd_fd->consumerd_fd;
> +
> +	ret = kernctl_snapshot_get_produced(infd, pos);
> +	if (ret != 0) {
> +		ret = errno;
> +		perror("kernctl_snapshot_get_produced");
> +	}
> +
> +	return ret;
> +}
> +
> +/*
>   * Poll on the should_quit pipe and the command socket return -1 on error and
>   * should exit, 0 if data is available on the command socket
>   */
> -- 
> 1.7.4.1
> 

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com