Messages in this thread | | | Date | Tue, 17 Jun 2008 22:57:54 +0800 | From | "Bob Zhang" <> | Subject | Re: Does anyone run AXFS in 2.4 version of kernel ? I am trying my best to apply AXFS patches into 2.4.18 |
| |
Today I carefully read the mkfs.axfs.c , I finally make clear about the axfs is how to organize the structure ,including the spec of super ,and the content of region , and how to store the file content in different region are .
Jared Hulber is a great programmer , this is a very complex thing . I am ready for writing a document about I have known for more newbies for axfs.
---- In fact , yesterday , I asked a stupid question to Jared Hulbert , the information from /proc/axfs/volume0 is engough to create a valid xml file .
for example ,from ./bin/busybox,0,18
./bin/busybox,28672,2
./bin/busybox,53248,4
./bin/busybox,143360,3
./bin/busybox,147456,3
./bin/busybox,151552,8
./bin/busybox,155648,11
we can know the page of offset 0 is called 18 times ,compared with other pages , it should be set XIP. <xipfsmap> <file> <name> ./bin/busybox </name> <chunk> <size>4096</size> <offset>0</offset> //maybe we need convert it from /proc info </chunk> </file> </xipfsmap> the advantage of AXFS is that we can set every page to be XIP , so the size is certainly 4096. but if we think like this , the size item in the xml is useless, I think maybe Jared Hulber want to provide some lazy method to set contiguous pages to be XIP , you don't need to write some chunk . but I feel that the chance of contiguous pages are read is not high , so for us , setting every page is enough .
if we want to set many pages to be XIP , we should write many chunk items .
Am I right , Mr Hulbert ?
Best Regards, Bob
| |