[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <cover.1674939002.git.nicolinc@nvidia.com>
Date: Sat, 28 Jan 2023 13:18:08 -0800
From: Nicolin Chen <nicolinc@...dia.com>
To: <jgg@...dia.com>, <kevin.tian@...el.com>
CC: <yi.l.liu@...el.com>, <iommu@...ts.linux.dev>,
<linux-kernel@...r.kernel.org>
Subject: [PATCH v2 0/3] iommufd: Remove iommufd_hw_pagetable_has_group
The iommufd_hw_pagetable_has_group is not a device-centric API and has
been a bit of a hack. And it needs to keep tracking an attached device
list on the hw_pagetable, and a device lock to protect the device list.
However, the coming domain replacement series can overcomplicate this
list/lock solution, especially to handle nested hw_pagetable use cases.
So, as a preparatory series, remove the device list/lock and also fix
the iommufd_hw_pagetable_has_group hack.
The iommufd_hw_pagetable_has_group() using the device list could be
replaced with a domain-pointer comparison between the hwpt->domain and
iommu_get_domain_for_dev(). The piece of dependency on list_empty() of
the device list can be also replaced with a refcount. Yet, the removal
of the device lock might introduce a race condition, so the ioas mutex
can be moved as an alternative protection.
You can also find this series on Github:
https://github.com/nicolinc/iommufd/commits/remove_iommufd_hw_pagetable_has_group
Changelog
v1->v2:
* Fixed a copy-n-paste mistake at a lockdep_assert_held() line
Thanks
Nicolin Chen
Nicolin Chen (2):
iommufd/device: Make hwpt_list list_add/del symmetric
iommufd/device: Change iommufd_hw_pagetable_has_group to device
centric
Yi Liu (1):
iommufd: Add devices_users to track the hw_pagetable usage by device
drivers/iommu/iommufd/device.c | 72 ++++++++++---------------
drivers/iommu/iommufd/hw_pagetable.c | 16 ++++--
drivers/iommu/iommufd/iommufd_private.h | 3 +-
3 files changed, 40 insertions(+), 51 deletions(-)
--
2.39.1
Powered by blists - more mailing lists