Christian R??ner via Postfix-devel:
> Hello,
>
> > Am 20.12.2024 um 19:26 schrieb Wietse Venema via Postfix-devel
> > <[email protected]>:
> >
> > Christian Roessner via Postfix-devel:
> >> Hi,
> >>
> >> why does exist a hard coded limit of 100000 characters in the
> >> netstring payload for socket maps? I have a customer who uses a
> >> socketmap for virtual aliases and he had to switch to the memcache
> >> protocol (he wrote a service speaking memcache) to workaround this
> >> issue, because memcache values seem not to have this limit.
> >
> > Even memcached has limits, he just hasn't run into them.
> >
> > Postfix virtual aliases have multiple limits. In fact, Postfix has
> > limits for everything, to prevent one bad actior from locking up
> > the entire server.
> >
> >> Would it be possible to either drop this limit,
> >
> > Sorry, that would make Postfix vulnerable by default, and that is
> > undesirable.
> >
> >> set it higher it or have a configuration parameter for it?
> >
> > It could be made configurable. For an example of this, see how the
> > "berkeley_db_read_buffer_size" configuration parameter controls a
> > global variable in the dict_db.c module.
> >
> > Such a global variable already exists in dict_sockmap.c.
>
> It is about 25 years ago that I did C-programming, so please be nice to me,
> if my first attempt is not perfect ;-)
>
> Could you please give me a hint, if the following approach goes into the
> right direction?
Very close; the dict_sockmap_max_reply variable can be owned by
the dict_sockmap class, and exported by the header file for that
class.
dict_sockmap.c: replace: int dict_sockmap_max_reply;
with: int dict_sockmap_max_reply = 100000;
(to test the dict_db.c code without linking it with
the rest of Postfix).
dict_sockmap.h: add: extern int dict_sockmap_max_reply;
mail_params.h: remove: extern int dict_sockmap_max_reply;
mail_params.c: add: #include <dict_sockmap.h>
mail_params.c: remove: int dict_sockmap_max_reply;
Wietse
_______________________________________________
Postfix-devel mailing list -- [email protected]
To unsubscribe send an email to [email protected]
