From: Christoffer Dall <christoffer.dall@linaro.org>
To: Antonios Motakis <a.motakis@virtualopensystems.com>
Cc: Alex Williamson <alex.williamson@redhat.com>,
kvm-arm <kvmarm@lists.cs.columbia.edu>,
Linux IOMMU <iommu@lists.linux-foundation.org>,
VirtualOpenSystems Technical Team <tech@virtualopensystems.com>,
alvise rigo <a.rigo@virtualopensystems.com>,
KVM devel mailing list <kvm@vger.kernel.org>,
Will Deacon <will.deacon@arm.com>,
Kim Phillips <kim.phillips@freescale.com>,
Stuart Yoder <stuart.yoder@freescale.com>,
Eric Auger <eric.auger@linaro.org>,
Catalin Marinas <catalin.marinas@arm.com>,
Mark Rutland <mark.rutland@arm.com>,
Vladimir Murzin <vladimir.murzin@arm.com>,
open list <linux-kernel@vger.kernel.org>
Subject: Re: [RFC PATCH v6 15/20] vfio/platform: support for maskable and automasked interrupts
Date: Wed, 10 Sep 2014 12:13:33 +0200 [thread overview]
Message-ID: <20140910101333.GA3002@lvm> (raw)
In-Reply-To: <CAG8rG2z02JPE+D-Bo1puuMPCR=wETciLaBgKT+i1XKQ55U-kYg@mail.gmail.com>
On Tue, Sep 02, 2014 at 06:06:17PM +0200, Antonios Motakis wrote:
> On Sun, Jun 8, 2014 at 12:17 PM, Christoffer Dall
> <christoffer.dall@linaro.org> wrote:
> > On Thu, Jun 05, 2014 at 07:03:23PM +0200, Antonios Motakis wrote:
> >> Adds support to mask interrupts, and also for automasked interrupts.
> >> Level sensitive interrupts are exposed as automasked interrupts and
> >> are masked and disabled automatically when they fire.
> >>
> >> Signed-off-by: Antonios Motakis <a.motakis@virtualopensystems.com>
> >> ---
> >> drivers/vfio/platform/vfio_platform_irq.c | 112 ++++++++++++++++++++++++--
> >> drivers/vfio/platform/vfio_platform_private.h | 2 +
> >> 2 files changed, 109 insertions(+), 5 deletions(-)
> >>
> >> diff --git a/drivers/vfio/platform/vfio_platform_irq.c b/drivers/vfio/platform/vfio_platform_irq.c
> >> index d79f5af..10dfbf0 100644
> >> --- a/drivers/vfio/platform/vfio_platform_irq.c
> >> +++ b/drivers/vfio/platform/vfio_platform_irq.c
> >> @@ -51,9 +51,17 @@ int vfio_platform_irq_init(struct vfio_platform_device *vdev)
> >> if (hwirq < 0)
> >> goto err;
> >>
> >> - vdev->irq[i].flags = VFIO_IRQ_INFO_EVENTFD;
> >> + spin_lock_init(&vdev->irq[i].lock);
> >> +
> >> + vdev->irq[i].flags = VFIO_IRQ_INFO_EVENTFD
> >> + | VFIO_IRQ_INFO_MASKABLE;
> >> +
> >> + if (irq_get_trigger_type(hwirq) & IRQ_TYPE_LEVEL_MASK)
> >> + vdev->irq[i].flags |= VFIO_IRQ_INFO_AUTOMASKED;
> >
> > This seems to rely on the fact that you had actually loaded a driver for
> > your device to set the right type. Is this assumption always correct?
> >
> > It seems to me that this configuration bit should now be up to your user
> > space drive who is the best candidate to know details about your device
> > at this point?
> >
>
> Hm, I see this type being set usually either in a device tree source,
> or in the support code for a specific platform. Are there any
> situations where this is actually set by the driver? If I understand
> right this is not the case for the PL330, but if it is possible that
> it is the case for another device then I need to rethink this. Though
> as far as I understand this should not be the case.
>
Wow, this has been incredibly long time since I looked at this code, so
not sure if I remember my original reasoning anymore, however,
while device properties are set in the DT, they would only be available
to this code if you actually loaded a device driver for that device,
right? I'm just not sure that assumption always holds for devices used
by VFIO, but I'm really not sure anymore. Maybe I'm rambling.
> >> +
> >> vdev->irq[i].count = 1;
> >> vdev->irq[i].hwirq = hwirq;
> >> + vdev->irq[i].masked = false;
> >> }
> >>
> >> vdev->num_irqs = cnt;
> >> @@ -77,11 +85,27 @@ void vfio_platform_irq_cleanup(struct vfio_platform_device *vdev)
> >>
> >> static irqreturn_t vfio_irq_handler(int irq, void *dev_id)
> >> {
> >> - struct eventfd_ctx *trigger = dev_id;
> >> + struct vfio_platform_irq *irq_ctx = dev_id;
> >> + unsigned long flags;
> >> + int ret = IRQ_NONE;
> >> +
> >> + spin_lock_irqsave(&irq_ctx->lock, flags);
> >> +
> >> + if (!irq_ctx->masked) {
> >> + ret = IRQ_HANDLED;
> >> +
> >> + if (irq_ctx->flags & VFIO_IRQ_INFO_AUTOMASKED) {
> >> + disable_irq_nosync(irq_ctx->hwirq);
> >> + irq_ctx->masked = true;
> >> + }
> >> + }
> >>
> >> - eventfd_signal(trigger, 1);
> >> + spin_unlock_irqrestore(&irq_ctx->lock, flags);
> >>
> >> - return IRQ_HANDLED;
> >> + if (ret == IRQ_HANDLED)
> >> + eventfd_signal(irq_ctx->trigger, 1);
> >> +
> >> + return ret;
> >> }
> >>
> >> static int vfio_set_trigger(struct vfio_platform_device *vdev,
> >> @@ -162,6 +186,82 @@ static int vfio_platform_set_irq_trigger(struct vfio_platform_device *vdev,
> >> return -EFAULT;
> >> }
> >>
> >> +static int vfio_platform_set_irq_unmask(struct vfio_platform_device *vdev,
> >> + unsigned index, unsigned start,
> >> + unsigned count, uint32_t flags, void *data)
> >> +{
> >> + uint8_t arr;
> >
> >
> > arr?
>
> arr for array! As in, the VFIO API allows an array of IRQs. However
> for platform devices we don't use this, each IRQ is exposed
> independently as an array of 1 IRQ.
>
but it's not an array. If it contains IRQs, call it irqs. Unless this
is referring specifically to a field *named* array, I don't remember the
API at current, but reading the code along it didn't make sense to me to
have a uint8_t called arr, and code should make as much sense
independenly as possible.
This reminds me of people writing code like:
String str;
yuck.
> >
> >> +
> >> + if (start != 0 || count != 1)
> >> + return -EINVAL;
> >> +
> >> + switch (flags & VFIO_IRQ_SET_DATA_TYPE_MASK) {
> >> + case VFIO_IRQ_SET_DATA_BOOL:
> >> + if (copy_from_user(&arr, data, sizeof(uint8_t)))
> >> + return -EFAULT;
> >> +
> >> + if (arr != 0x1)
> >> + return -EINVAL;
> >
> > why the fallthrough, what's this about?
>
> The VFIO API allows to unmask/mask an array of IRQs, however with
> platform devices we only have arrays of 1 IRQ (so not really arrays).
>
> So if the user uses VFIO_IRQ_SET_DATA_BOOL, we need to check that arr
> == 0x1. When that is the case, a fallthrough to the same code for
> VFIO_IRQ_SET_DATA_NONE is safe.
>
> If that is not readable enough, then I can add a comment or duplicate
> the code that does the unmasking. I realize that if you don't know the
> VFIO API well, then this can look confusing.
>
yeah, please put a big fat comment explaining the fallthrough.
-Christoffer
next prev parent reply other threads:[~2014-09-10 10:13 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <1401987808-23596-1-git-send-email-a.motakis@virtualopensystems.com>
2014-06-05 17:03 ` [RFC PATCH v6 01/20] iommu/arm-smmu: change IOMMU_EXEC to IOMMU_NOEXEC Antonios Motakis
2014-06-16 15:04 ` Will Deacon
2014-06-05 17:03 ` [RFC PATCH v6 02/20] iommu: add capability IOMMU_CAP_NOEXEC Antonios Motakis
2014-06-05 20:03 ` Alex Williamson
2014-06-05 17:03 ` [RFC PATCH v6 03/20] iommu/arm-smmu: add IOMMU_CAP_NOEXEC to the ARM SMMU driver Antonios Motakis
2014-06-16 15:04 ` Will Deacon
2014-06-16 15:25 ` Alex Williamson
2014-06-16 15:30 ` Will Deacon
2014-06-05 17:03 ` [RFC PATCH v6 04/20] iommu/arm-smmu: add capability IOMMU_CAP_INTR_REMAP Antonios Motakis
2014-06-05 18:31 ` Varun Sethi
2014-06-08 10:31 ` Christoffer Dall
2014-06-16 14:53 ` Joerg Roedel
2014-06-16 15:13 ` Will Deacon
2014-06-16 15:21 ` Joerg Roedel
2014-06-16 15:25 ` Will Deacon
2014-06-16 15:38 ` Joerg Roedel
2014-06-26 18:08 ` Chalamarla, Tirumalesh
2014-06-26 18:15 ` Chalamarla, Tirumalesh
2014-06-26 18:41 ` Chalamarla, Tirumalesh
2014-06-26 19:00 ` Alex Williamson
2014-06-26 19:10 ` Chalamarla, Tirumalesh
2014-06-26 19:36 ` Alex Williamson
2014-06-27 8:47 ` Will Deacon
2014-06-27 21:57 ` Chalamarla, Tirumalesh
2014-06-28 7:05 ` Marc Zyngier
2014-06-16 15:30 ` Alex Williamson
2014-06-05 17:03 ` [RFC PATCH v6 05/20] vfio/iommu_type1: support for platform bus devices on ARM Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 06/20] vfio: introduce the VFIO_DMA_MAP_FLAG_NOEXEC flag Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 07/20] vfio/iommu_type1: implement " Antonios Motakis
2014-06-05 20:48 ` Alex Williamson
2014-06-05 17:03 ` [RFC PATCH v6 08/20] driver core: platform: add device binding path 'driver_override' Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 09/20] vfio/platform: initial skeleton of VFIO support for platform devices Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 10/20] vfio/platform: return info for device and its memory mapped IO regions Antonios Motakis
2014-06-05 21:14 ` Alex Williamson
2014-06-06 16:39 ` Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 11/20] vfio/platform: read and write support for the device fd Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 12/20] vfio/platform: support MMAP of MMIO regions Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 13/20] vfio/platform: return IRQ info Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 14/20] vfio/platform: initial interrupts support Antonios Motakis
2014-06-08 10:09 ` Christoffer Dall
2014-09-02 16:07 ` Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 15/20] vfio/platform: support for maskable and automasked interrupts Antonios Motakis
2014-06-08 10:17 ` Christoffer Dall
2014-09-02 16:06 ` Antonios Motakis
2014-09-10 10:13 ` Christoffer Dall [this message]
2014-09-11 17:20 ` Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 16/20] vfio: move eventfd support code for VFIO_PCI to a sepparate file Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 17/20] vfio: add local lock in virqfd instead of depending on VFIO PCI Antonios Motakis
2014-06-05 22:19 ` Alex Williamson
2014-06-06 16:57 ` Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 18/20] vfio: pass an opaque pointer on virqfd initialization Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 19/20] vfio: initialize the virqfd workqueue in VFIO generic code Antonios Motakis
2014-06-05 17:03 ` [RFC PATCH v6 20/20] vfio/platform: implement IRQ masking/unmasking via an eventfd Antonios Motakis
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20140910101333.GA3002@lvm \
--to=christoffer.dall@linaro.org \
--cc=a.motakis@virtualopensystems.com \
--cc=a.rigo@virtualopensystems.com \
--cc=alex.williamson@redhat.com \
--cc=catalin.marinas@arm.com \
--cc=eric.auger@linaro.org \
--cc=iommu@lists.linux-foundation.org \
--cc=kim.phillips@freescale.com \
--cc=kvm@vger.kernel.org \
--cc=kvmarm@lists.cs.columbia.edu \
--cc=linux-kernel@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=stuart.yoder@freescale.com \
--cc=tech@virtualopensystems.com \
--cc=vladimir.murzin@arm.com \
--cc=will.deacon@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).