Re: How to keep the kernel from crashing on cd9660 error ?

["Thomas Schmitt" <scdbackup%gmx.net@localhost>]

Hi,

> EOPNOTSUPP doesn't look right here ...

The file is complying to ISO 9660 specs but indigestible
for any reader that relies on block-to-block mapping.
One would need byte-to-byte mapping to support that file.
NetBSD cannot, and it seems Linux isofs can't either.

Do you have an idea for a better errno in such a case ?


> The upcoming conversion to vcache should make it easier to deal with
> these problems.  I will add some priority to the conversion of cd9660.

Priority for cd9660 is highly appreciated, of course.

I have pending kern/48808 for mount option -s, and i am fine-tuning
the proposal for large data files.
So we would need to coordinate.

It would be helpful, if a test could be made with the error
generator code in my initial mail of this thread.
I have looked into other fs drivers of NetBSD meanwhile.
They all do similar error handling.

So i have to consider the possibility that only my NetBSD
copy is faulty.
The bug fixes, which i proposed and got comitted, can hardly
be to blame for any memory problem.

All is well, as long as i do not want to bail out after having
allocated the vnode. But the kernel becomes a mine field if
i do not follow the code path up to the end of function
cd9660_vget_internal().
Does this ring a bell ?


Have a nice day :)

Thomas