On Feb 10, 2004, at 8:29 AM, Andrew Brown wrote:

> On Tue, Feb 10, 2004 at 08:01:45AM -0800, Matt Thomas wrote:
>>> ...
>>> how can asserting UVM_FLAG_NOMERGE prevent splitting unless the
>>> allocation is less than three pages in size?  can't i simply make a
>>> three page allocation with UVM_FLAG_NOMERGE set and then unmap the
>>> middle page?
>>
>> True but the current uses of uvm_map and uvm_unmap in the kernel
>> don't do that.
>>
>> They simply undo a previous uvm_map.
>
> well, okay then.  :)
>
> so how many kernel map entries do you have today?

matt@amd64> pmap 0
FFFF800000000000  21972K read/write/exec     [ anon ]
FFFF800001575000   4372K read/write/exec     [ uvm_aobj ]
FFFF8000019BA000 131072K read/write/exec     [ kmem_map ]
FFFF8000099BA000  16384K read/write/exec     [ pager_map ]
FFFF80000A9BA000   9756K read/write/exec     [ uvm_aobj ]
FFFF80000B341000   4096K read/write/exec     [ exec_map ]
FFFF80000B741000   1200K read/write/exec     [ phys_map ]
FFFF80000B86D000   4096K read/write/exec     [ mb_map ]
FFFF80000BC6D000   5764K read/write/exec     [ uvm_aobj ]
FFFF80000C20E000   8192K read/write/exec     [ ubc_pager ]
FFFF80000CA0E000  22092K read/write/exec     [ uvm_aobj ]
FFFF80000DFAD000     76K read/write/exec     [ uvm_aobj ]
FFFF80000DFC4000  61196K read/write/exec     [ uvm_aobj ]
FFFF800011B8F000  46940K read/write/exec     [ uvm_aobj ]
FFFF800014970000  48184K read/write/exec     [ uvm_aobj ]
FFFF800017880000  15880K read/write/exec     [ uvm_aobj ]
FFFF800018810000  32192K read/write/exec     [ uvm_aobj ]
  total           433464K


-- 
Matt Thomas                     email: matt@3am-software.com
3am Software Foundry              www: http://3am-software.com/bio/matt/
Cupertino, CA              disclaimer: I avow all knowledge of this 
message.