path: root/tip22/README
diff options
authorGuido Guenther <agx@sigxcpu.org>2006-11-18 23:48:07 +0100
committerGuido Guenther <agx@bogon.sigxcpu.org>2006-11-18 23:48:07 +0100
commitd1e063beb43e595680c65e3804d1f8ddff53373b (patch)
treef7256dfe1b807920270ec5113df6f6e4abf1ed0f /tip22/README
Imported Debian version
Diffstat (limited to 'tip22/README')
1 files changed, 32 insertions, 0 deletions
diff --git a/tip22/README b/tip22/README
new file mode 100644
index 0000000..becc313
--- /dev/null
+++ b/tip22/README
@@ -0,0 +1,32 @@
+How to create the boot image:
+tip22 <vmlinux> <initrd> <outputfile>
+tip32 <vmlinux> <initrd> <outputfile>
+- vmlinux is an ELF Linux kernel. For versions before 2.6.9 you need to
+ apply an extra patch for correct commandline parsing, see
+ /usr/share/doc/tip22/examples.
+- initrd is the gzip compressed initial ramdisk you want add
+- outputfile is the bootimage. It is in ECOFF format for ip22 and in
+ ELF format for ip32. You can for example tfpboot this image on an SGI
+ Indy using the bootp(): command in the prom.
+it basically works like this:
+- tftpload.o is a relocatable object created during package build time
+- the tip22 script converts the ramdisk and vmlinux to relocatable
+ objects using linker scripts including symbols that mark the begin
+ and the end of kernel and ramdisk
+- those are linked together into the output file using yet another
+ linker script
+- when executed by the prom the embedded loader:
+ - memcopies the kernel to its loadaddress
+ - memcopies the initrd after the kernel to a page boundary
+ - parses the command line and everything within an append="..."
+ statement is passed on to the kernel, e.g.
+ 'bootp(): append="console=ttyS0"'
+ - jumps into the kernel passing the start address and size of the
+ initrd on the command line
+ -- Thu, 15 May 2002 23:07:18 +0200 Guido Guenther <agx@sigxcpu.org>