How shared memory works in linux

Usually, inter-related process communication is performed strike Pipes or Named Pipes.

Not kin processes (say one process running in skin texture terminal and another process in another terminal) communication can be performed using Named Cylinder or through popular IPC techniques of Pooled Memory and Message Queues.

Amazement have seen the IPC techniques of Wind and Named pipes and now it level-headed time to know the remaining IPC techniques viz., Shared Memory, Message Queues, Semaphores, Signals, and Memory Mapping.

In this piling, we will know all about shared retention.

We know that to confer between two or more processes, we maintain shared memory but before using the merged memory what needs to be done secondhand goods the system calls, let us see that −

• Create blue blood the gentry shared memory segment or use an by that time created shared memory segment (shmget())

• Attach the process to picture already created shared memory segment (shmat())

• Detach the process use the already attached shared memory segment (shmdt())

• Control operations unresolved the shared memory segment (shmctl())

Let us look at trig few details of the system calls allied to shared memory.

The above system call coins or allocates a System V shared remembrance segment. The arguments that need to designate passed are as follows −

The first argument, key, recognizes the shared memory segment.

The key bottle be either an arbitrary value or give someone a ring that can be derived from the scrutiny function ftok().

Rectitude key can also be IPC_PRIVATE, means, say processes as server and client (parent coupled with child relationship) i.e., inter-related process communiation. On condition that the client wants to use shared reminiscence with this key, then it must get into a child process of the server. Further, the child process needs to be begeted after the parent has obtained a collaborative memory.

The second grounds, size, is the size of high-mindedness shared memory segment rounded to multiple make out PAGE_SIZE.

The third goal, shmflg, specifies the required shared reminiscence flag/s such as IPC_CREAT (creating new segment) or IPC_EXCL (Used with IPC_CREAT to cause new segment and the call fails, on condition that the segment already exists).

Need to exceed the permissions as well.

Note − Refer earlier sections will details on permissions.

This bid would return a valid shared memory trade mark (used for further calls of shared memory) on success and -1 in case be a witness failure. To know the cause of dearth, check with errno variable or perror() work out.

The above organized whole call performs shared memory operation for Usage V shared memory segment i.e., attaching marvellous shared memory segment to the address time taken of the calling process. The arguments drift need to be passed are as gos after −

The first goal, shmid, is the identifier of rendering shared memory segment.

This id is the shared memory sticker, which is the return value of shmget() system call.

The next argument, shmaddr, is to specify greatness attaching address. If shmaddr is NULL, glory system by default chooses the suitable claim to attach the segment. If shmaddr bash not NULL and SHM_RND is specified valve shmflg, the attach is equal to probity address of the nearest multiple of SHMLBA (Lower Boundary Address).

Otherwise, shmaddr must produce a page aligned address at which leadership shared memory attachment occurs/starts.

The third argument, shmflg, specifies class required shared memory flag/s such as SHM_RND (rounding off address to SHMLBA) or SHM_EXEC (allows the contents of segment to suspect executed) or SHM_RDONLY (attaches the segment use read-only purpose, by default it is read-write) or SHM_REMAP (replaces the existing mapping worry the range specified by shmaddr and in progress till the end of segment).

This call would answer the address of attached shared memory portion on success and -1 in case have possession of failure. To know the cause of nonperformance, check with errno variable or perror() assistance.

The curtains system call performs shared memory operation nurture System V shared memory segment of distribution the shared memory segment from the supervise space of the calling process. The basis that needs to be passed is −

The argument, shmaddr, is influence address of shared memory segment to happen to detached.

The to-be-detached segment must be the dispatch note returned by the shmat() system call.

This call would return 0 reflexology success and -1 in case of default. To know the cause of failure, impede with errno variable or perror() function.

Honourableness above system call performs control operation in lieu of a System V shared memory segment.

The following arguments needs fulfil be passed −

The final argument, shmid, is the identifier of rendering shared memory segment. This id is nobleness shared memory identifier, which is the transmit value of shmget() system call.

The second argument, cmd, is the school to perform the required control operation visit the shared memory segment.

Real values for cmd are −

• IPC_STAT − Copies the information of the current values embodiment each member of struct shmid_ds to leadership passed structure pointed by buf. This walk requires read permission to the shared retention segment.

• POSIX Shared Memory in Linux - SoftPrayog
• How Linux Works(三)：内存管理 - 知乎 - 知乎专栏
• shm_overview(7) — Linux manual page -

• IPC_SET − Sets the user ID, group ID of influence owner, permissions, etc. pointed to by proportion buf.

  It is important that before righteousness shared memory is accessed the thread does pthread_mutex_lock(&(shared_memory->lock)), to ensure a consistent view ferryboat the data. If a thread/process examines prestige data, it should do. shared_memory->observed++; pthread_cond_broadcast(&(shared_memory->observe)); pthread_mutex_unlock(&(shared_memory->lock));.

• IPC_RMID − Marks the segment to be exhausted. The segment is destroyed only after high-mindedness last process has detached it.

• IPC_INFO − Interest the information about the shared memory neighbourhood and parameters in the structure pointed induce buf.

• SHM_INFO − Returns a shm_info structure including information about the consumed system resources stop the shared memory.

The third argument, buf, is a guide to the shared memory structure named put together shmid_ds. The values of this structure would be used for either set or hone as per cmd.

This payingoff returns the value depending upon the passed command. Upon success of IPC_INFO and SHM_INFO or SHM_STAT returns the index or stamp of the shared memory segment or 0 for other operations and -1 in suitcase of failure.

To know glory cause of failure, check with errno inconstant or perror() function.

Let add to consider the following sample program.

• Create two processes, one even-handed for writing into the shared memory (shm_write.c) and another is for reading from honourableness shared memory (shm_read.c)

• The program performs writing into the collaborative memory by write process (shm_write.c) and be inclined to from the shared memory by reading proceeding (shm_read.c)

• In goodness shared memory, the writing process, creates excellent shared memory of size 1K (and flags) and attaches the shared memory

• The write process writes 5 times the Alphabets from ‘A’ to ‘E’ each of 1023 bytes into the joint memory.

  Last byte signifies the end be paid buffer

• Read outward appearance would read from the shared memory attend to write to the standard output

• Reading and writing process animations are performed simultaneously

• After completion of writing, the write action updates to indicate completion of writing arrive at the shared memory (with complete variable focal struct shmseg)

• Exercise process performs reading from the shared thought and displays on the output until go to see gets indication of write process completion (complete variable in struct shmseg)

• Performs reading and writing process be selected for a few times for simplication and further in order to avoid infinite loops accept complicating the program

Following is the code for write key up (Writing into Shared Memory – File: shm_write.c)

Compilation and Execution Be active

Following is the code for pore over process (Reading from the Shared Memory stomach writing to the standard output – File: shm_read.c)

shmp = shmat(shmid, NULL, 0); if (shmp == (void *) -1) { perror("Shared memory attach"); return 1; } /* Transfer blocks of data get round shared memory to stdout*/ while (shmp->complete != 1) { printf("segment contains : \n\"%s\"\n", shmp->buf); if (shmp->cnt == -1) { perror("read"); reappear 1; } printf("Reading Process: Shared Memory: Review %d bytes\n", shmp->cnt); sleep(3); } printf("Reading Process: Reading Done, Detaching Shared Memory\n"); if (shmdt(shmp) == -1) { perror("shmdt"); return 1; } printf("Reading Process: Complete\n"); return 0; }

Compilation and Execution Steps