A lot of memory pages allocated by kalloc() (HP-UX 11.00)

[Jdamian]

Hi

My HP-UX 11.00 box shows a high utilisation of memory near 100%.

The most strange is the high use of dynamic memory pages allocated by kalloc(), displayed by kmeminfo (866481 pages from 1170108):

Physical memory usage summary (in pages):

Physmem = 3670016 Available physical memory:
Freemem = 108262 Free physical memory
Used = 3561754 Used physical memory:
System = 1541998 by kernel:
Static = 71050 for text and static data
Dynamic = 1170108 for dynamic data
Bufcache = 293601 for file-system buffer cache
Eqmem = 71 for equiv. mapped page pool
SCmem = 7168 for system critical page pool
User = 1699865 by user processes
Uarea = 7032 for thread uareas
Disowned = 88258 disowned pages

----------------------------------------------------------------------
Dynamic memory usage summary (in pages):

Dynamic = 1170108 Kernel dynamic data (sysmap):
MALLOC = 303518 memory buckets:
bucket[ 5] = 5772 size 32 bytes
bucket[ 6] = 264 size 64 bytes
bucket[ 7] = 14852 size 128 bytes
bucket[ 8] = 1638 size 256 bytes
bucket[ 9] = 2683 size 512 bytes
bucket[10] = 1969 size 1024 bytes
bucket[11] = 261580 size 2048 bytes
bucket[12] = 1150 size 4096 bytes
bucket[13] = 346 size 2 pages
bucket[14] = 417 size 3 pages
bucket[15] = 124 size 4 pages
bucket[16] = 630 size 5 pages
bucket[17] = 48 size 6 pages
bucket[18] = 70 size 7 pages
bucket[19] = 480 size 8 pages
bucket[20] = 11495 size > 8 pages
Kalloc = 866481 kalloc()
Eqalloc = 95 eqalloc()
Reserved = 14 Reserved pools

Some questions:

1) Does kalloc() mean "kernel allocation" ? What is it?

2) How to reduce the number of such type?

3) Is that a normal behaviour?


Thanx in advance

[whiteknight]

hi

Some questions:

1) Does kalloc() mean "kernel allocation" ? What is it?
usually we get kalloc() out of virtual memory space.
Sysmap is a resource map used by the kernel to manage free areas of virtual space available to the kernel. The map is a fixed size (in pages) and address pairs. As kernel virtual space is allocated, sysmap is updated appropriately.

Sysmap records the size and address of free areas of virtual space. As areas are freed and so returned to sysmap, the kernel will coalesce the returned entry with an existing entry, if possible. If the area being returned is not adjacent to an already free area, which cannot be coalesced with an existing entry, the kernel adds a new entry to sysmap. As this happens more often, which it will as a normal occurrence, the number of entries in sysmap grows. This is the "fragmentation" we are referring to.



2) How to reduce the number of such type?
1 â There are some known problems with memory leaks that can cause this fragmentation/panic. Below you will find a list of patches that address known memory leak issues and should be installed (if applicable) on any system having sysmap problems.

NOTE: These patches are current as of this publishing but may have been superceded since.
NOTE: The below list does not include dependencies and will need to be checked.
NOTE: Listed below is a general memory leak patch list. It includes specific hardware leak issues. Please be sure hardware specific patches apply to your system.

PHKL_25906 s700_800 11.00 Probe,sysproc,IDDS,PM,VM,PA-8700,asyncio
PHNE_23833 s700_800 11.00 ONC/NFS General Release/Performance Patch
PHNE_22159 s700_800 11.00 telnet kernel and telnetd(1M) patch
PHCO_23966 s700_800 11.00 Software Distributor (SD) Cumulative Patch
PHKL_24004 s700_800 11.00 SCSI IO Subsystem Cumulative Patch
PHNE_22566 s700_800 11.00 Cumulative STREAMS Patch
PHNE_24413 s700_800 11.00 LAN product cumulative patch
PHNE_23456 s700_800 11.00 cumulative ARPA Transport patch
PHNE_24100 s700_800 11.00 Streams Pty cumulative patch
PHKL_22208 s700_800 11.00 Fix VxFS DMAPI DPFs and Memory Leak
PHKL_23409 s700_800 11.00 NFS, Large Data Space, kernel memory leak
PHKL_18543 s700_800 11.00 PM/VM/UFS/async/scsi/io/DMAPI/JFS/perf patch



Hardware Specific:


PHNE_21390 s700_800 11.00 HSC FDDI B.11.00.0[0-7] cumulative patch
PHNE_23768 s700_800 11.00 EISA FDDI B.11.00.0[0-7] cumulative patch
PHNE_20949 s700_800 11.00 HyperFabric B.11.00.0[0-8] cumulative patch
PHNE_24025 s700_800 11.00 HP-PB FDDI (J2157B) product cumulative patch
PHNE_15537 s700_800 11.00 Fibre Channel cumulative patch


3) Is that a normal behaviour?
To me, I will fix the known issue first.

WK