sash commands
I've had a busy evening implementing the hpux syscalls required to make some of the sash commands work. The following commands work:
cd
-chgrp
-chmod
-chown
help
-ls
-mkdir
-pwd
-touch
-where
-mount is known to not work. I need to speak to prumpf about why this is. -umount probably works, but without -mount, that's hard to test :-) Ditto -sync. Maybe some others work. If there's any sash command you particularly want to work which doesn't yet, drop me a mail.
many thanks to Grant for all his help this evening.
Signature · 4 lines
Matthew Wilcox <[email protected]> "Windows and MacOS are products, contrived by engineers in the service of specific companies. Unix, by contrast, is not so much a product as it is a painstakingly compiled oral history of the hacker subculture." - N Stephenson
Review
Quoted 10 lines
cd
(quoted diff) +0
-3-chgrp -chmod -chownhelp
(quoted diff) +0
-5-ls -mkdir -pwd -touch -where
whee.
Quoted 1 line
-mount is known to not work. I need to speak to prumpf about why this is.
If I see this correctly this is the 4-argument problem. Our current (hacked) syscall support only supports the register arguments because supporting the other ones too just would mean nobody ever got around to fix syscall support :)
If anyone else needs to send me email (or needed in the last two weeks or so), please use the new ugly address [email protected]. As I don't work for SuSE any more, [email protected] is very likely to be invalid shortly and I am not sure how often I'll be able to check it before.
Quoted 1 line
Ditto -sync.
what's the problem with sync ?
Philipp Rumpf
Quoted 5 lines · Philipp Rumpf
Quoted 1 line
-mount is known to not work. I need to speak to prumpf about why this is.
If I see this correctly this is the 4-argument problem. Our current (hacked) syscall support only supports the register arguments because supporting the other ones too just would mean nobody ever got around to fix syscall support :)
Partly. I should be less gnomic..
Yes, the 4-argument problem is relevant, but it's not the only thing. sash (since it's a linux program) uses the linux arguments to sys_mount(), so rather than both writing hpux_mount(), I decided to convince sash to use the Linux entry point for sys_mount(), and that's what the sash-willy in puffin.external.hp.com's /tmp has different to the other sashes.
Then I discovered that Linux syscalls were not working because we didn't have a page mapped at 0xC000'1000, then I discovered we didn't have a page mapped at 0xC000'0000 either and there was instead a nasty kludge in fault.c to handle hpux syscalls. So I added a nasty kludge in there to make linux syscalls work too (and they don't have the 4-argument limitation, due to the syscall convention which I decreed).
Now I'm blundering around in the mm system trying to figure out how to get gateway pages working properly.
Quoted 4 lines
If anyone else needs to send me email (or needed in the last two weeks or so), please use the new ugly address [email protected]. As I don't work for SuSE any more, [email protected] is very likely to be invalid shortly and I am not sure how often I'll be able to check it before.
The return address on this one was [email protected]. I'm using the ugly address you quote there.
Quoted 3 lines
Quoted 1 line
Ditto -sync.
what's the problem with sync ?
The same as the problem with -umount -- rather hard to test that it's working :-)
Signature · 4 lines
Matthew Wilcox <[email protected]> "Windows and MacOS are products, contrived by engineers in the service of specific companies. Unix, by contrast, is not so much a product as it is a painstakingly compiled oral history of the hacker subculture." - N Stephenson
Quoted 11 lines
Quoted 3 lines
If I see this correctly this is the 4-argument problem. Our current (hacked) syscall support only supports the register arguments because supporting the other ones too just would mean nobody ever got around to fix syscall support :)
Partly. I should be less gnomic..
Yes, the 4-argument problem is relevant, but it's not the only thing. sash (since it's a linux program) uses the linux arguments to
sys_mount(), so rather than both writinghpux_mount(), I decided to convince sash to use the Linux entry point forsys_mount(), and that's what the sash-willy in puffin.external.hp.com's /tmp has different to the other sashes.
Oh. I was under the impression mount was POSIX-defined (it is not). Sure, getting it to use Linux syscalls is the best thing to do.
Quoted 9 lines
Then I discovered that Linux syscalls were not working because we didn't have a page mapped at 0xC000'1000, then I discovered we didn't have a page mapped at 0xC000'0000 either and there was instead a nasty kludge in
fault.cto handle hpux syscalls. So I added a nasty kludge in there to make linux syscalls work too (and they don't have the 4-argument limitation, due to the syscall convention which I decreed).Now I'm blundering around in the mm system trying to figure out how to get gateway pages working properly.
Well, the first step would be to get PAGE_OFFSET == 0x8000 0000 (or any other value that does not collide with 0xc000 0000) working. I tried to keep dependencies on it at a minimum but completely failed, so if someone could find all instances of
depi 3,1,2,XXX and replace them with tovirt XXX0xc0000000 (as PAGE_OFFSET) PAGE_OFFSET / __PAGE_OFFSET
that would help a lot (otherwise I'll try).
Philipp Rumpf