Re: [PATCH v5 4/6] 9pfs: fix wrong errno being sent to Linux client on macOS host

[Christian Schoenebeck <qemu_oss@crudebyte.com>]

On Freitag, 29. April 2022 15:29:15 CEST Greg Kurz wrote:
> On Fri, 29 Apr 2022 21:19:51 +0800
> 
> Bin Meng <bmeng.cn@gmail.com> wrote:
> > On Fri, Apr 29, 2022 at 9:08 PM Greg Kurz <groug@kaod.org> wrote:
> > > On Fri, 29 Apr 2022 14:46:26 +0200
> > > 
> > > Christian Schoenebeck <qemu_oss@crudebyte.com> wrote:
> > > > On Freitag, 29. April 2022 13:28:39 CEST Bin Meng wrote:
> > > > > On Fri, Apr 29, 2022 at 7:16 PM Christian Schoenebeck
> > > > > 
> > > > > <qemu_oss@crudebyte.com> wrote:
> > > > > > Linux and macOS only share some errno definitions with equal macro
> > > > > > name and value. In fact most mappings for errno are completely
> > > > > > different on the two systems.
> > > > > > 
> > > > > > This patch converts some important errno values from macOS host to
> > > > > > corresponding Linux errno values before eventually sending such
> > > > > > error
> > > > > > codes along with 'Rlerror' replies (if 9p2000.L is used that is).
> > > > > > Not
> > > > > > having translated errnos before violated the 9p2000.L protocol
> > > > > > spec,
> > > > > > 
> > > > > > which says:
> > > > > >   "
> > > > > >   size[4] Rlerror tag[2] ecode[4]
> > > > > >   
> > > > > >   ... ecode is a numerical Linux errno.
> > > > > >   "
> > > > > >   
> > > > > >   https://github.com/chaos/diod/wiki/protocol#lerror----return-err
> > > > > >   or-code
> > > > > > 
> > > > > > This patch fixes a bunch of misbehaviours when running a Linux
> > > > > > client
> > > > > > 
> > > > > > on macOS host. For instance this patch fixes:
> > > > > >   mount -t 9p -o posixacl ...
> > > > > > 
> > > > > > on Linux guest if security_mode=mapped was used for 9p server,
> > > > > > which
> > > > > > refused to mount successfully, because macOS returned ENOATTR==93
> > > > > > when client tried to retrieve POSIX ACL xattrs, because errno 93
> > > > > > is defined as EPROTONOSUPPORT==93 on Linux, so Linux client
> > > > > > believed
> > > > > > that xattrs were not supported by filesystem on host in general.
> > > > > 
> > > > > This issue looks exact the same issue we were trying to fix when
> > > > > supporting 9p on Windows host,
> > > > > 
> > > > > What we did is like this:
> > > > > http://patchwork.ozlabs.org/project/qemu-devel/patch/20220425142705.
> > > > > 2099270-> 10-bmeng.cn@gmail.com/
> > > > > 
> > > > > But we had some questions in mind (see the commit message of our
> > > > > patch, and below)
> > > > > 
> > > > > > Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
> > > > > > Link:
> > > > > > https://lore.kernel.org/qemu-devel/20220421124835.3e664669@bahia/
> > > > > > Reviewed-by: Greg Kurz <groug@kaod.org>
> > > > > > ---
> > > > > > 
> > > > > >  hw/9pfs/9p-util.h | 30 ++++++++++++++++++++++++++++++
> > > > > >  hw/9pfs/9p.c      |  2 ++
> > > > > >  2 files changed, 32 insertions(+)
> > > > > > 
> > > > > > diff --git a/hw/9pfs/9p-util.h b/hw/9pfs/9p-util.h
> > > > > > index 2cc9a5dbfb..c3526144c9 100644
> > > > > > --- a/hw/9pfs/9p-util.h
> > > > > > +++ b/hw/9pfs/9p-util.h
> > > > > > @@ -58,6 +58,36 @@ static inline uint64_t
> > > > > > host_dev_to_dotl_dev(dev_t dev)
> > > > > > 
> > > > > >  #endif
> > > > > >  }
> > > > > > 
> > > > > > +/* Translates errno from host -> Linux if needed */
> > > > > > +static inline int errno_to_dotl(int err) {
> > > > > > +#if defined(CONFIG_LINUX)
> > > > > > +    /* nothing to translate (Linux -> Linux) */
> > > > > > +#elif defined(CONFIG_DARWIN)
> > > > > > +    /*
> > > > > > +     * translation mandatory for macOS hosts
> > > > > > +     *
> > > > > > +     * FIXME: Only most important errnos translated here yet,
> > > > > > this should
> > > > > > be +     * extended to as many errnos being translated as possible
> > > > > > in
> > > > > > future. +     */
> > > > > > +    if (err == ENAMETOOLONG) {
> > > > > > +        err = 36; /* ==ENAMETOOLONG on Linux */
> > > > > > +    } else if (err == ENOTEMPTY) {
> > > > > > +        err = 39; /* ==ENOTEMPTY on Linux */
> > > > > > +    } else if (err == ELOOP) {
> > > > > > +        err = 40; /* ==ELOOP on Linux */
> > > > > > +    } else if (err == ENOATTR) {
> > > > > > +        err = 61; /* ==ENODATA on Linux */
> > > > > > +    } else if (err == ENOTSUP) {
> > > > > > +        err = 95; /* ==EOPNOTSUPP on Linux */
> > > > > > +    } else if (err == EOPNOTSUPP) {
> > > > > > +        err = 95; /* ==EOPNOTSUPP on Linux */
> > > > > > +    }
> > > > > 
> > > > > What happens if a macOS guest is running on QEMU from a macOS host?
> > > > > Here all macOS errnos are translated to the Linux errnos. Will macOS
> > > > > be happy?
> > > > 
> > > > Look at the commit log of this patch: it is a matter of which protocol
> > > > is used> > > 
> > > > (currently there are 3 [1] protocol versions):
> > > >    * 9p2000: nothing to translate here, as this protocol version does
> > > >    not
> > > >    
> > > >      return a numeric error code, it only returns an error string (and
> > > >      we are
> > > >      no longer supporting 9p2000 version in QEMU anyway BTW [1]):
> > > >      http://ericvh.github.io/9p-rfc/rfc9p2000.html#anchor27
> > > >    
> > > >    * 9p2000.L: errno *must* be in Linux errno mapping:
> > > >      https://github.com/chaos/diod/wiki/protocol#lerror----return-erro
> > > >      r-code
> > > >    
> > > >    * 9p2000.u: this one returns both an error code and error string,
> > > >    and it
> > > >    
> > > >      says the error string should be preferred being interpreted by
> > > >      client:
> > > >      http://ericvh.github.io/9p-rfc/rfc9p2000.u.html#anchor15
> > > > 
> > > > In this patch here I only translated errno for 9p2000.L, whereas you
> > > > are
> > > > always translating it, no matter wich protocol version is used. You
> > > > might
> > > > argue that there should be a translation for 9p2000.u as well, but in
> > > > the end we don't know the OS running on guest in this case. It could
> > > > be Linux or something else.
> > > 
> > > In the case of 9p2000.u the spec says "to provide a hint of the
> > > underlying
> > > UNIX error number which caused the error on the server" and even
> > > mentions
> > > "consistency problems of mapping error numbers betweeen different
> > > versions
> > > of UNIX"... this basically means that errno in 9p2000.u is undefined
> > > since
> > > it depends on the host. It is thus unusable unless the guest runs a
> > > compatible UNIX variant. In any case, there's really nothing to
> > > translate.
> > > 
> > > > [1] https://wiki.qemu.org/Documentation/9p#9p_Protocol
> > 
> > Thanks for the clarifications and pointers to different protocols! It
> > looks what we did in our Windows patch is correct.

Like I said, you are translating it for all protocol version, whereas this 
patch here translates errnos for 9p2000.L version only.

> > I have another question, does this mean the current 9pfs client on
> > macOS is broken since it does not use any translation? With this
> > patch, now the 9p server returns the translated linux errno so the 9p
> > client on macOS should complain.
> 
> I don't now the macOS client but if it doesn't expect linux errnos
> then it is infringing 9p2000.L and should be fixed.

Agreed, if you are using 9p2000.L with that macOS client and client does not 
translate errnos Linux -> macOS, then client is broken. In the end it matters 
what the protocol documentation specified.

Which client is that? Is it from Apple or from a third party? And are you sure 
you were actually using 9p2000.L and not 9p2000.u?

Best regards,
Christian Schoenebeck