Re: [RFC PATCH v10 1/5] iommu/vt-d: add pci_dev parameter to qi_submit_sync and refactor callers

From: Ethan Zhao
Date: Wed Jan 10 2024 - 02:52:10 EST



On 1/10/2024 12:59 PM, Baolu Lu wrote:
On 12/29/23 1:02 AM, Ethan Zhao wrote:
Signed-off-by: Ethan Zhao<haifeng.zhao@xxxxxxxxxxxxxxx>

Please don't leave the message body empty. You should describe why do
you want to add the change in this patch.
Seems the description part was lost, will append next version.

---
  drivers/iommu/intel/dmar.c          | 45 +++++++++++++++++++++--------
  drivers/iommu/intel/iommu.c         | 26 +++++------------
  drivers/iommu/intel/iommu.h         | 17 +++++------
  drivers/iommu/intel/irq_remapping.c |  2 +-
  drivers/iommu/intel/pasid.c         | 11 ++-----
  drivers/iommu/intel/svm.c           | 13 ++++-----
  6 files changed, 58 insertions(+), 56 deletions(-)

diff --git a/drivers/iommu/intel/dmar.c b/drivers/iommu/intel/dmar.c
index 23cb80d62a9a..3d661f2b7946 100644
--- a/drivers/iommu/intel/dmar.c
+++ b/drivers/iommu/intel/dmar.c
@@ -1344,7 +1344,7 @@ static int qi_check_fault(struct intel_iommu *iommu, int index, int wait_index)
   * can be part of the submission but it will not be polled for completion.
   */
  int qi_submit_sync(struct intel_iommu *iommu, struct qi_desc *desc,
-           unsigned int count, unsigned long options)
+           unsigned int count, unsigned long options, struct pci_dev *pdev)

How about adding a bit in options parameter to tell whether the @pdev is
valid?

well, checking the option bit or checking pdev == NULL, use one parameter

to describe another one is common function defination method if one

parameter couldn't self-describe.

This case, we always check pdev(one checking), and if we check option bit first, then have

to check pdev again (one or two checking).  I prefer checking pdev only.


Thanks,

Ethan


Best regards,
baolu