[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160609181102.13455.33593.stgit@bhelgaas-glaptop2.roam.corp.google.com>
Date: Thu, 09 Jun 2016 13:20:08 -0500
From: Bjorn Helgaas <bhelgaas@...gle.com>
To: Yinghai Lu <yinghai@...nel.org>,
Benjamin Herrenschmidt <benh@...nel.crashing.org>,
Michal Simek <monstr@...str.eu>,
Paul Mackerras <paulus@...ba.org>,
Michael Ellerman <mpe@...erman.id.au>
Cc: linux-pci@...r.kernel.org, linuxppc-dev@...ts.ozlabs.org,
linux-kernel@...r.kernel.org
Subject: [PATCH v1 0/3] PCI: Clean up write combining usage for user mmap
These patches:
- Enforce a "no write combining on I/O port space mapping" policy.
This seems like an obviously good thing but was previously
enforced only by powerpc.
- Stop giving powerpc users write combining mappings of prefetchable
memory they ask for write combining with the procfs
PCIIOC_WRITE_COMBINE ioctl or a sysfs "resourceN_wc" file.
- Clean up arch-specific code that related to the above.
Ben, I'm particularly interested in your thoughts about the powerpc
change. It should "only" affect performance of users who aren't
actually requesting write combining. I don't know how many of those
users there might be.
Yinghai, I added your sign-off to the powerpc part, since you posted
that exact patch. Let me know if you don't want that.
These are on my pci/resource branch.
---
Bjorn Helgaas (2):
PCI: Ignore write combining when mapping I/O port space
microblaze/PCI: Remove useless __pci_mmap_set_pgprot()
Yinghai Lu (1):
powerpc/pci: Remove __pci_mmap_set_pgprot()
arch/microblaze/pci/pci-common.c | 31 +------------------------------
arch/powerpc/kernel/pci-common.c | 37 ++++---------------------------------
drivers/pci/proc.c | 9 ++++++---
3 files changed, 11 insertions(+), 66 deletions(-)
Powered by blists - more mailing lists