From: Marc Zyngier <marc.zyngier@arm.com>
To: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Yingjoe Chen <yingjoe.chen@mediatek.com>,
Thomas Gleixner <tglx@linutronix.de>,
Mark Rutland <Mark.Rutland@arm.com>,
Boris BREZILLON <boris.brezillon@free-electrons.com>,
Russell King <linux@arm.linux.org.uk>,
Jason Cooper <jason@lakedaemon.net>,
Pawel Moll <Pawel.Moll@arm.com>,
"devicetree\@vger.kernel.org" <devicetree@vger.kernel.org>,
"hc.yen\@mediatek.com" <hc.yen@mediatek.com>,
"srv_heupstream\@mediatek.com" <srv_heupstream@mediatek.com>,
"yh.chen\@mediatek.com" <yh.chen@mediatek.com>,
"linux-kernel\@vger.kernel.org" <linux-kernel@vger.kernel.org>,
"grant.likely\@linaro.org" <grant.likely@linaro.org>,
Yijing Wang <wangyijing@huawei.com>,
Rob Herring <robh+dt@kernel.org>,
"nathan.chung\@mediatek.com" <nathan.chung@mediatek.com>,
"yingjoe.chen\@gmail.com" <yingjoe.chen@gmail.com>,
Matthias Brugger <matthias.bgg@gmail.com>,
"eddie.huang\@mediatek.com" <eddie.huang@mediatek.com>,
Bjorn Helgaas <bhelgaas@google.com>,
Sascha Hauer <kernel@pengutronix.de>,
"lin ux- arm-kernel\@lists.infradead.org"
<linux-arm-kernel@lists.infradead.org>
Subject: Re: [PATCH v7 1/4] irqchip: gic: Support hierarchy irq domain.
Date: Thu, 20 Nov 2014 10:07:21 +0000 [thread overview]
Message-ID: <87wq6q9odi.fsf@approximate.cambridge.arm.com> (raw)
In-Reply-To: <546D6D62.4030005@linux.intel.com> (Jiang Liu's message of "Thu, 20 Nov 2014 04:26:10 +0000")
On Thu, Nov 20 2014 at 4:26:10 am GMT, Jiang Liu <jiang.liu@linux.intel.com> wrote:
Hi Jiang,
> On 2014/11/20 1:18, Marc Zyngier wrote:
>> Hi Yingjoe,
>>
>> On Wed, Nov 19 2014 at 2:14:08 pm GMT, Yingjoe Chen
>> <yingjoe.chen@mediatek.com> wrote:
>>> +
>>> +static const struct irq_domain_ops gic_irq_domain_hierarchy_ops = {
>>> + .xlate = gic_irq_domain_xlate,
>>> + .alloc = gic_irq_domain_alloc,
>>> + .free = irq_domain_free_irqs_top,
>>
>> I'm convinced that irq_domain_free_irqs_top is the wrong function to
>> call here, because you're calling it from the bottom, not the top-level
>> (it has no parent).
>>
>> I cannot verify this with your code as I don't a working platform with
>> GICv2m, but if I enable something similar on GICv3, it dies a very
>> painful way:
>>
>> Unable to handle kernel NULL pointer dereference at virtual address 00000018
>> pgd = ffffffc03d059000
>> [00000018] *pgd=0000000081356003, *pud=0000000081356003, *pmd=0000000000000000
>> Internal error: Oops: 96000006 [#1] SMP
>> Modules linked in:
>> CPU: 4 PID: 1052 Comm: sh Not tainted 3.18.0-rc4+ #3311
>> task: ffffffc03e320000 ti: ffffffc001390000 task.ti: ffffffc001390000
>> PC is at irq_domain_free_irqs_recursive+0x1c/0x80
>> LR is at irq_domain_free_irqs_common+0x88/0x9c
>> pc : [<ffffffc0000ed790>] lr : [<ffffffc0000ede20>] pstate: 60000145
>> [...]
>> [<ffffffc0000ed790>] irq_domain_free_irqs_recursive+0x1c/0x80
>> [<ffffffc0000ede1c>] irq_domain_free_irqs_common+0x84/0x9c
>> [<ffffffc0000ede98>] irq_domain_free_irqs_top+0x64/0x7c <-- gic_domain.free()
>> [<ffffffc0000ed798>] irq_domain_free_irqs_recursive+0x24/0x80
>> [<ffffffc0000ee468>] irq_domain_free_irqs_parent+0x14/0x20
>> [<ffffffc0003500b8>] its_irq_domain_free+0xc8/0x250
>> [<ffffffc0000ed798>] irq_domain_free_irqs_recursive+0x24/0x80
>> [<ffffffc0000ede1c>] irq_domain_free_irqs_common+0x84/0x9c
>> [<ffffffc0000ede98>] irq_domain_free_irqs_top+0x64/0x7c
>> [<ffffffc0000ef518>] msi_domain_free+0x70/0x88
>> [<ffffffc0000ed798>] irq_domain_free_irqs_recursive+0x24/0x80
>> [<ffffffc0000ee3ac>] irq_domain_free_irqs+0x108/0x17c
>> [<ffffffc0000efb68>] msi_domain_free_irqs+0x28/0x4c
>> [<ffffffc000369cac>] free_msi_irqs+0xb4/0x1c0
>> [<ffffffc00036adec>] pci_disable_msix+0x3c/0x4c
>> [...]
>>
>> and I cannot see how this could work on the standard GIC either.
>>
>> Thomas, Jiang: could you please confirm or infirm my suspicions? My
>> understanding is that irq_domain_free_irqs_top can only be called from
>> the top-level domain.
> Hi Marc,
> It indicates that irq_domain_free_irqs_top() is not a good name.
> We have:
> 1) irq_domain_set_hwirq_and_chip() to set irq_chip and chip_data
> 2) irq_domain_set_info() to set irq_chip, chip_data, flow_handler and
> handler_data;
> 3) irq_domain_reset_irq_data() resets irq_chip and chip_data.
> 4) irq_domain_free_irqs_common() resets irq_chip, chip_data and calls
> parent domain's domain_ops.free() callback.
> 5) irq_domain_free_irqs_top() resets irq_chip, chip_data, flow handler,
> handler_data and call parent domain's domain_ops.free() callback.
Yes, and this "call parent domain's free callback" is where the problem
lies. Here, it is called from the innermost domain, with no parent.
> So there two possible improvements here:
> 1) Rename irq_domain_free_irqs_top() with better name, any suggestions?
> It's named as is because it's always called by the outer-most
> irqdomains on x86.
> 2) Change irq_domain_free_irqs_common() and irq_domain_free_irqs_top()
> to call parent domain's domain_ops.free() callback only if parent
> exists. By this way, they could be used for inner-most irqdomains.
> If OK, I will respin a version 4 patch set based on tip/irq/irqdomain.
> Thoughts?
Checking the parent is probably a safe solution (this is not a hot path
anyway). I don't care much about the name though, and I the only thing I
can think of is irq_domain_free_irqs_reset_flow, which looks so bad it's
not even funny. I'll let the matter rest in your capable hands! ;-)
Thanks,
M.
--
Jazz is not dead. It just smells funny.
next prev parent reply other threads:[~2014-11-20 10:07 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-11-19 14:14 [PATCH v7 0/4] ARM: mediatek: Add support for interrupt polarity Yingjoe Chen
2014-11-19 14:14 ` [PATCH v7 1/4] irqchip: gic: Support hierarchy irq domain Yingjoe Chen
2014-11-19 17:18 ` Marc Zyngier
2014-11-20 3:57 ` Yingjoe Chen
2014-11-20 9:41 ` Yingjoe Chen
2014-11-20 10:29 ` Marc Zyngier
2014-11-20 4:26 ` Jiang Liu
2014-11-20 10:07 ` Marc Zyngier [this message]
2014-11-21 15:51 ` Yingjoe Chen
2014-11-24 19:31 ` Marc Zyngier
2014-11-19 14:14 ` [PATCH v7 2/4] ARM: mediatek: Add sysirq interrupt polarity support Yingjoe Chen
2014-11-19 18:04 ` Mark Rutland
2014-11-21 15:36 ` Yingjoe Chen
2014-11-19 14:14 ` [PATCH v7 3/4] ARM: mediatek: Add sysirq in mt6589/mt8135/mt8127 dtsi Yingjoe Chen
2014-11-19 17:49 ` Marc Zyngier
2014-11-20 3:29 ` Yingjoe Chen
2014-11-19 14:14 ` [PATCH v7 4/4] dt-bindings: add bindings for mediatek sysirq Yingjoe Chen
2014-11-19 18:07 ` Mark Rutland
2014-11-24 15:14 ` Yingjoe Chen
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=87wq6q9odi.fsf@approximate.cambridge.arm.com \
--to=marc.zyngier@arm.com \
--cc=Mark.Rutland@arm.com \
--cc=Pawel.Moll@arm.com \
--cc=bhelgaas@google.com \
--cc=boris.brezillon@free-electrons.com \
--cc=devicetree@vger.kernel.org \
--cc=eddie.huang@mediatek.com \
--cc=grant.likely@linaro.org \
--cc=hc.yen@mediatek.com \
--cc=jason@lakedaemon.net \
--cc=jiang.liu@linux.intel.com \
--cc=kernel@pengutronix.de \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux@arm.linux.org.uk \
--cc=matthias.bgg@gmail.com \
--cc=nathan.chung@mediatek.com \
--cc=robh+dt@kernel.org \
--cc=srv_heupstream@mediatek.com \
--cc=tglx@linutronix.de \
--cc=wangyijing@huawei.com \
--cc=yh.chen@mediatek.com \
--cc=yingjoe.chen@gmail.com \
--cc=yingjoe.chen@mediatek.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).