brintos

brintos / linux-shallow public Read only

0
0
Text · 3.7 KiB · 8e296bc Raw
107 lines · plain
1.. SPDX-License-Identifier: GPL-2.02 3==================================4Tracefs ring-buffer memory mapping5==================================6 7:Author: Vincent Donnefort <vdonnefort@google.com>8 9Overview10========11Tracefs ring-buffer memory map provides an efficient method to stream data12as no memory copy is necessary. The application mapping the ring-buffer becomes13then a consumer for that ring-buffer, in a similar fashion to trace_pipe.14 15Memory mapping setup16====================17The mapping works with a mmap() of the trace_pipe_raw interface.18 19The first system page of the mapping contains ring-buffer statistics and20description. It is referred to as the meta-page. One of the most important21fields of the meta-page is the reader. It contains the sub-buffer ID which can22be safely read by the mapper (see ring-buffer-design.rst).23 24The meta-page is followed by all the sub-buffers, ordered by ascending ID. It is25therefore effortless to know where the reader starts in the mapping:26 27.. code-block:: c28 29        reader_id = meta->reader->id;30        reader_offset = meta->meta_page_size + reader_id * meta->subbuf_size;31 32When the application is done with the current reader, it can get a new one using33the trace_pipe_raw ioctl() TRACE_MMAP_IOCTL_GET_READER. This ioctl also updates34the meta-page fields.35 36Limitations37===========38When a mapping is in place on a Tracefs ring-buffer, it is not possible to39either resize it (either by increasing the entire size of the ring-buffer or40each subbuf). It is also not possible to use snapshot and causes splice to copy41the ring buffer data instead of using the copyless swap from the ring buffer.42 43Concurrent readers (either another application mapping that ring-buffer or the44kernel with trace_pipe) are allowed but not recommended. They will compete for45the ring-buffer and the output is unpredictable, just like concurrent readers on46trace_pipe would be.47 48Example49=======50 51.. code-block:: c52 53        #include <fcntl.h>54        #include <stdio.h>55        #include <stdlib.h>56        #include <unistd.h>57 58        #include <linux/trace_mmap.h>59 60        #include <sys/mman.h>61        #include <sys/ioctl.h>62 63        #define TRACE_PIPE_RAW "/sys/kernel/tracing/per_cpu/cpu0/trace_pipe_raw"64 65        int main(void)66        {67                int page_size = getpagesize(), fd, reader_id;68                unsigned long meta_len, data_len;69                struct trace_buffer_meta *meta;70                void *map, *reader, *data;71 72                fd = open(TRACE_PIPE_RAW, O_RDONLY | O_NONBLOCK);73                if (fd < 0)74                        exit(EXIT_FAILURE);75 76                map = mmap(NULL, page_size, PROT_READ, MAP_SHARED, fd, 0);77                if (map == MAP_FAILED)78                        exit(EXIT_FAILURE);79 80                meta = (struct trace_buffer_meta *)map;81                meta_len = meta->meta_page_size;82 83                printf("entries:        %llu\n", meta->entries);84                printf("overrun:        %llu\n", meta->overrun);85                printf("read:           %llu\n", meta->read);86                printf("nr_subbufs:     %u\n", meta->nr_subbufs);87 88                data_len = meta->subbuf_size * meta->nr_subbufs;89                data = mmap(NULL, data_len, PROT_READ, MAP_SHARED, fd, meta_len);90                if (data == MAP_FAILED)91                        exit(EXIT_FAILURE);92 93                if (ioctl(fd, TRACE_MMAP_IOCTL_GET_READER) < 0)94                        exit(EXIT_FAILURE);95 96                reader_id = meta->reader.id;97                reader = data + meta->subbuf_size * reader_id;98 99                printf("Current reader address: %p\n", reader);100 101                munmap(data, data_len);102                munmap(meta, meta_len);103                close (fd);104 105                return 0;106        }107