>Number:         24857
>Category:       kern
>Synopsis:       panic: uvm_mapent_alloc: out of static map entries, check MAX_KMAPENT (currently 1114)
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sat Mar 20 06:26:00 UTC 2004
>Closed-Date:
>Last-Modified:
>Originator:     Simon Burge
>Release:        NetBSD 1.6ZK, sources from March 4, 2004.
>Organization:
Wasabi Systems
>Environment:
	System: NetBSD bigkev.thistledown.com.au 1.6ZK NetBSD 1.6ZK (BIGKEV)
		#21: Thu Mar 4 09:35:17 EST 2004
		simonb@bigkev.thistledown.com.au:/var/tmp/BIGKEV i386
	Architecture: i386
	Machine: i386

	Dual Xeon with 2GB RAM, most filesystems on raidframe mirrors.
	The kernel config is GENERIC.MP and the following additions:

		makeoptions     DEBUG="-g"
		options         DIAGNOSTIC
		options         LOCKDEBUG
		options         MPDEBUG
		options         MPVERBOSE
		options         DEBUG

		options         KMEMSTATS

		options         MSGBUFSIZE=131072

		options         MPACPI
		options         ACPIVERBOSE
		acpi0           at mainbus0

>Description:
	An SMP box has been struggling to stay up for week recently.

	reboot    ~                         Sat Mar 20 16:48
	reboot    ~                         Mon Mar 15 12:59
	reboot    ~                         Tue Mar 09 10:34
	reboot    ~                         Thu Mar 04 09:56 (new kernel)

	Usually it remains running, but creating a pipe fails - a symptom
	traced back to no more KVA available.  Usually the box hasn't been
	able to create a crash dump.

	The most recent event was a panic:

	panic: uvm_mapent_alloc: out of static map entries, check MAX_KMAPENT (currently 1114)
	Stopped in pid 21.1 (pagedaemon) at     netbsd:cpu_Debugger+0x4:        leave

	This time I was able to get a crash dump.  "vmstat -m" says:

		Memory totals:  In Use    Free    Requests
				13815K   1990K    50591506

		In use 416222K, total allocated 603028K; utilization 69.0%

	which is well down on some earlier figures taken while the system
	was running - for example:

		Memory totals:  In Use    Free    Requests
				13682K   1112K     5965223

		In use 751043K, total allocated 816072K; utilization 92.0%

	"pmap -p 0"  n the core file shows:

		C0000000  54576K read/write/exec     [ anon ]
		C354C000    136K read/write/exec     [ uvm_aobj ]
		C356E000 131072K read/write/exec     [ kmem_map ]
		CB56E000  16384K read/write/exec     [ pager_map ]
		CC56E000  23000K read/write/exec     [ uvm_aobj ]
		CDBE4000   4096K read/write/exec     [ exec_map ]
		CDFE4000   1200K read/write/exec     [ phys_map ]
		CE110000   2048K read/write/exec     [ mb_map ]
		CE310000   1160K read/write/exec     [ uvm_aobj ]
		CE432000   8192K read/write/exec     [ ubc_pager ]
		CEC32000  32844K read/write/exec     [ uvm_aobj ]
		D0C46000   2192K read/write/exec     [ uvm_aobj ]
			[ 7812 more lines ]
		 total   834332K 

	which might be a bit more worrying.

	I still have the crash dump if any more info is needed.

>How-To-Repeat:
	Run an SMP box for a while?

>Fix:
	Non given.
>Release-Note:
>Audit-Trail:
>Unformatted: