On a Monday in 2023, Michal Prívozník wrote:
On 9/11/23 15:51, Ján Tomko wrote:
>
https://bugzilla.redhat.com/show_bug.cgi?id=2034630
>
https://gitlab.com/libvirt/libvirt/-/issues/535
>
https://gitlab.gnome.org/GNOME/gnome-boxes/-/issues/292
>
> Ján Tomko (8):
> qemu: fix indentation
> conf: move idmap definition earlier
> conf: move idmap parsing earlier
> conf: add idmap element to filesystem
> qemu: format uid/gid map for virtiofs
> qemu: virtiofs: do not force UID 0
> qemu: allow running virtiofsd in session mode
> docs: virtiofs: add section about ID remapping
>
> docs/formatdomain.rst | 7 +
> docs/kbase/virtiofs.rst | 29 ++++
> src/conf/domain_conf.c | 149 ++++++++++++------
> src/conf/domain_conf.h | 29 ++--
> src/conf/schemas/domaincommon.rng | 3 +
> src/qemu/qemu_validate.c | 25 ++-
> src/qemu/qemu_virtiofs.c | 17 +-
> .../vhost-user-fs-fd-memory.xml | 4 +
> 8 files changed, 181 insertions(+), 82 deletions(-)
>
So how does this idmap end up on virtiofsd cmd line? Maybe you forgot to
send some additional patches?
It's in PATCH 5/8: qemu: format uid/gid map for virtiofs
Perhaps you asking this question is a good indicator that I should add
some qemuvirtiofsxml2argvtest.
However, in the linked bug, German suggested that the user namespace
be created by libvirt, because they intend to remove that --uid-map
option from virtiofsd, so the virtiofsd command line change will be
missing from v2 of this series.
Jano
Michal