Unix & Linux Stack Exchange is a question and answer site for users of Linux, FreeBSD and other Un*x-like operating systems. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I have a large tar file (60GB) containing image files. I'm using mmap() on this entire file to read in these images, which are accessed randomly.

I'm using mmap() for the following reasons:

  1. Thread safety -- I cannot seek an ifstream from multiple threads.
  2. I can avoid extra buffering.
  3. I get some caching (in the form of a requested page already being resident.)

The question is what happens when I've read every image in that 60GB file? Certainly not all of the images are being used at once -- they're read, displayed, and then discarded.

My mmap() call is:

mmap(0, totalSize, PROT_READ, MAP_SHARED | MAP_NORESERVE, fd, 0); 

Here's the question: does the kernel see that I've mapped read-only pages backed by a file and simply purges the unused pages on memory pressure? I'm not sure if this case is recognized. Man pages indicate that MAP_NORESERVE will not require backing swap space, but there doesn't seem to be any guarantee of what happens to the pages under memory pressure. Is there any guarantee that the kernel will purge my unneeded pages before it, say, purges the filesystem cache or OOM's another process?


share|improve this question

migrated from stackoverflow.com Sep 14 '11 at 10:01

This question came from our site for professional and enthusiast programmers.

The kernel knows that your pages are ephemeral. Only the source of the kernel knows if it will make intelligent use of that information. Google 'swappiness' for a headache. – bmargulies Sep 14 '11 at 0:43
You can use madvise to give it "advice" on what you're doing with the memory. I'm not sure how well it makes use of your advice, but a MADV_DONTNEED on regions you know you're done with at least won't hurt anything. – Scott Lamb Sep 14 '11 at 1:58
up vote 2 down vote accepted

A read-only mmap is largely equivalent to open followed by lseek and read. If a chunk of memory that's mapped in a process is backed up by a file, the copy in RAM is considered part of the disk cache, and will be freed under memory pressure, just like a disk cache entry created by reading from a file.

I haven't checked the source, but I believe MAP_NORESERVE makes no difference for read-only mappings.

share|improve this answer
You say that a page mapped in in this way is counted as part of the disk cache -- do you have a source for that? Thanks! – Devin Lane Sep 19 '11 at 11:41

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.