$ grep ^pgfault /proc/vmstat
pgfault 9687532485 17 days: 6,600 a second
$ ps -eo rss= | awk '{ s += $1 } END { print s / 1024 " MB" }'
58339 MB on a machine with 61 GB, 41 in use
$ swapon --show
/dev/dm-1 partition 80G 834.2M memory that is not in memory
Lecture 04: physical memory arrives in a page fault. This lecture: what happens inside one.
The story
Two views of one address space
$ grep -c . /proc/$$/maps 33 zones
pages: 2647 present: 1098 from VmSize and VmRSS
Two edits
VSZ RSS MINFL
after mmap(4 MB) 6848 1692 64 the list
after writing byte 0 of every page 6848 5788 1088 the table, one fault each
after writing byte 4095 of every page 6848 5788 1088 nothing
Inside a page fault
Four verdicts
verdict
the address is
the frame
the process sees
minor fault
in a zone, allowed
found in memory
nothing, ~1 µs
major fault
in a zone, allowed
read from disk
a delay, ms
SEGV_MAPERR
in no zone
–
SIGSEGV
SEGV_ACCERR
in a zone that forbids it
–
SIGSEGV
$ ./segfault 6 SIGSEGV at 0x7dd401855000: SEGV_ACCERR: the zone forbids this access
$ ./segfault 3 SIGSEGV at 0x755bca209000: SEGV_MAPERR: no zone covers the address
$ ./major_fault first pass (from disk): 16384 major faults, 0.302 s
second pass (from cache): 1024 minor faults, 0.002 s
Where frames come from
Swap, in a box
$ systemd-run --user --scope -p MemoryMax=64M ./swap
start: RSS 1 MB swap 0 MB major faults 0 0.00 s
after writing 256 MB: RSS 63 MB swap 193 MB major faults 3 0.62 s
after reading it back: RSS 64 MB swap 192 MB major faults 8424 2.19 s
Without the box: 257 MB resident, 0 major faults, 0.00 s.
The page fault is where the kernel does its memory work.
It decides, from the list of zones, between fixing and killing; it brings frames from zero, from files and from swap; it shares frames between processes and copies them only on a write.
The page table makes it affordable: a tree, so that it fits, and a TLB, so that it is rarely read.
