I've read some articles about Linux memory management and understand (perhaps not proficient) the concept of free memory, available memory, swap, etc. But when I check with the following command:

cat /proc/meminfo; free -w; vmstat -s

I just can't calculate the numbers precisely to prove my understanding.

For example, "cat /proc/meminfo" shows:

MemTotal:        8169968 kB
MemFree:          984124 kB
MemAvailable:    4231224 kB
Buffers:          171420 kB
Cached:          3272688 kB
SwapCached:        65604 kB
Active:          3415076 kB
Inactive:        3388312 kB
Active(anon):    2636088 kB
Inactive(anon):   789932 kB
Active(file):     778988 kB
Inactive(file):  2598380 kB
Unevictable:          64 kB

"free -w" shows:

              total        used        free      shared     buffers       cache   available
Mem:        8169968     3566616      983864       66952      171420     3448068     4230964
Swap:       3903484      500620     3402864

"vmstat -s" shows

  8169968 K total memory
  3566616 K used memory
  3415476 K active memory
  3388312 K inactive memory
   983864 K free memory
   171420 K buffer memory
  3448068 K swap cache
  3903484 K total swap
   500620 K used swap
  3402864 K free swap
  9335996 non-nice user cpu ticks

And here are my questions:

1. Why doesn't "used + available = total"?

Some articles say that "free" memory is not used by anything, while "available" memory might be used by, for example some buffer, but can be allocated to programes at any time. As per my understanding, it means:

total = available + used

But when I calculate:

4231224 ("MemAvailable" in /proc/meminfo) + 3566616 ("used memory" in vmstat) = 7797840

Gap to "total" (8169968 K): 372128 K (or 363 M) less

4230964 ("available" in free) + 3566616 ("used" in free) = 7797580

Gap to "total" (8169968 K): 372388 K (or 363 M) less

So why is there a large gap?

2. Why doesn't "free + buffer + cache = available"?

According to some articles, Linux may use some memory for buffer and cache, so this part of memory are not "free" (not used by anything, so it can be allocated to programes immediately) but still "available" (despite being used for buffer or cache, but can be freed at any time and allocated to programes). But when I calculate:


984124 (MemFree) + 171420 (Buffers) + 3272688 (Cached) = 4428232

Gap to "MemAvailable" (4231224): 197008 K (or 192 M) more


983864 (free) + 171420 (buffers) + 3448068 (cache) = 4603352

Gap to "available" (4230964): 372388 K (or 363 M) more

Why is there a large gap?

Furthermore, number "372388 K" occured in both this and the last question. That is, in "free" command:

total = available + used + 372388
total = free + buffers + cache - 372388

So what is it (the "372388 K" memory)?

3. Why doesn't "active + inactive + free = total"?

Some articles say that "active" is memory accessed recently, while "inactive" is that not. As per my understanding:

active (application) + inactive (application) = memory used by applications active (buffer) + inactive (buffer) = memory used by buffer ...

So "active + inactive = total used memory" and "total used memory + free = total memory". But when I calculate:


3415076 (Active) + 3388312 (Inactive) + 984124 (MemFree) = 7787512

Gap to total (8169968 K): 382456 K (or 373 M) less


3415476 (active) + 3388312 (inactive) + 983864 (free) = 7787652

Gap to total (8169968 K): 382316 K (or 373 M) less

Is my understanding correct? Are there any other similar equations?

The reason I'm asking this question is not for writing an application to calculate memory, but rather for validating my understanding and finding out what I missed.

For example if I was told that my salary is $1000 per month but got just $800 on my bank accout. Some articles say that the reason is tax. I calculated 12% tax but there is still $80 gap, so I asked again and then got to know the concept of social insurance.

Output of those Linux memory statistics is such a case: except that "free swap + used swap = total swap" can be precisely calculated, I can't get any equation from the output. This makes me unconfident with my understanding of Linux memory management. I think I've missed something, especially when the gap is up to 373 MB.

1 Answer 1

  1. Why doesn't "used + available = total"?

Because available memory is a subset of used and free memory: it’s the amount of memory which can be used without forcing the system to swap, i.e. free memory plus however much memory is currently used for cache which can be evicted immediately. See How can I get the amount of available memory portably across distributions? for details of the calculation.

  1. Why doesn't "free + buffer + cache = available"?

Because some of the memory used for buffers and cache isn’t available: it can’t be evicted immediately, or using it would push the system above a watermark.

  1. Why doesn't "active + inactive + free = total"?

Because the active and inactive values reflect the corresponding LRUs, and some pages aren’t tracked there — for example, page tables (PageTables) — or are unevictable (Unevictable).

Are there any other similar equations?

In free’s output, “used” is calculated as “total – free – cached – buffers”, unless the result is negative (which can happen e.g. in LXC) in which case it’s “total – free”. Thus in all cases, “total” is either “used + free + cached + buffers”, or “used + free”.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .