Re: [PATCH 0/1] Rewrite /hurd/symlink on top of trivfs OFF TOPIC PRAISE

[jbranso]

June 21, 2021 9:22 AM, "Sergey Bugaev" <bugaevc@gmail.com> wrote:

> Hello,
> 
> I came by this issue [0], which basically states that /hurd/symlink was
> behaving in a weird way. Namely, when opened with O_NOLINK, it would
> appear to be an empty file, not a symlink. The report talks about broken
> symlinks, but the issue is more fundamental:

Thanks for working on this!  I'm wanting to learn more about trivfs for my 
caesar
cipher eventually, but looking that how to write trivial translators makes
me a little curious...I'll learn more about it someday.  :)

It would be cool to have a gemini translator.  It would only take a few 100 
lines of code.

> 
> $ settrans -ac /tmp/demo /hurd/symlink /etc/hostname
> $ cat /tmp/demo
> sergey-hurd-box
> $ file /tmp/demo
> /tmp/demo: empty
> $ ls -l /tmp/demo
> -rw-r--r-- 1 sergey sergey 0 Jun 21 15:52 /tmp/demo
> $ fsysopts /tmp/demo
> ext2fs --writable --relatime --no-inherit-dir-group --store-type=typed 
> device:hd0s1
> 
> [0]: https://www.gnu.org/software/hurd/open_issues/symlink_translator.html
> 
> I took a look at trans/symlink.c, and found out that:
> 
> * It was not using libtrivs, as I would expect it to.
> 
> * It was treating O_NOLINK like O_NOTRANS; namely by returning the underlying
> node. This is the cause of the issue: O_NOLINK has very different semantics.
> It should open the link itself, and then suppport io_stat () and io_read ()
> on it, returning S_IFLNK | 0777 for the mode, and the link target as
> contents.
> 
> * It was leaking ports (should deallocate dotdot when not MOVE_SEND'ing it),
> and skipping some error checks.
> 
> So I thought I'd try rewriting /hurd/symlink to use trivfs. This is possible
> because trivfs has this convenient trivfs_getroot_hook mechanism, which lets 
> me
> either handle fsys_getroot () any way I want to, or continue with the default
> implementation found in trivfs (for O_NOLINK).
> 
> End result: it seems to work great. Absolute, relative; file, ls, ls -l, stat,
> readlink, cat; everything just does what I'd expect it to. Even fsysopts now
> shows the invocation properly. Some demos:
> 
> $ settrans -ac /tmp/demo trans/symlink /etc/hostname
> $ cat /etc/hostname
> sergey-hurd-box
> $ cat /tmp/demo
> sergey-hurd-box
> $ file /tmp/demo
> /tmp/demo: symbolic link to /etc/hostname
> $ ls -l /tmp/demo
> lrwxrwxrwx 1 sergey sergey 13 Jun 21 16:11 /tmp/demo -> /etc/hostname
> $ fsysopts /tmp/demo
> trans/symlink /etc/hostname
> 
> $ settrans -ac /tmp/demo trans/symlink no-such-file
> $ cat /tmp/demo
> cat: /tmp/demo: No such file or directory
> $ file /tmp/demo
> /tmp/demo: broken symbolic link to no-such-file
> $ ls -l /tmp/demo
> lrwxrwxrwx 1 sergey sergey 12 Jun 21 16:11 /tmp/demo -> no-such-file
> 
> Sergey Bugaev (1):
> trans/symlink.c: Rewrite on top of trivfs
> 
> trans/Makefile | 10 +-
> trans/symlink.c | 243 +++++++++++++++++++++++++++---------------------
> 2 files changed, 137 insertions(+), 116 deletions(-)
> 
> -- 
> 2.31.1