> On Aug 4, 2020, at 5:24 PM, Samuel Thibault wrote:
>
> Re,
>
> Junling Ma, le dim. 02 août 2020 21:07:00 -0700, a ecrit:
>> +struct pt_regs;
>> +extern int request_irq (unsigned int irq, void (*handler) (int, void *,
>> struct pt_regs *),
>> +
> On Aug 4, 2020, at 5:30 PM, Samuel Thibault wrote:
>
> Junling Ma, le mar. 04 août 2020 17:08:03 -0700, a ecrit:
>>>> static user_intr_t *
>>>> search_intr (struct irqdev *dev, ipc_port_t dst_port)
>>>> {
>>>> user_intr_t *e;
>&g
In the next patch, we disallow deliver_user_intr from touching the queue, so we
do not need to lock it. In fact, by doing that, we prevent interrupts from
pulling the rug under our feet, so there is no need to do splhigh/splx in
irq_acknowledge and intr_thread.
Junling
> On Aug 4, 2020, at 4:5
Hi Samual,
> On Aug 4, 2020, at 4:51 PM, Samuel Thibault wrote:
>
> Junling Ma, le dim. 02 août 2020 21:04:24 -0700, a ecrit:
>> -struct irqdev irqtab = {
>> - "irq", irq_eoi, &main_intr_queue, 0,
>> - {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,
these two device_* called
that are only meant for the irq device.
Best,
Junling
> On Aug 4, 2020, at 3:58 PM, Samuel Thibault wrote:
>
> Junling Ma, le mar. 04 août 2020 14:47:09 -0700, a ecrit:
>> The tot_num_intr field is a count of how many deliverable interrupts across
>&g
Hi Jess,
> On Aug 4, 2020, at 3:22 PM, Jessica Clarke wrote:
>
> On 4 Aug 2020, at 22:47, Junling Ma wrote:
>>
>> The tot_num_intr field is a count of how many deliverable interrupts across
>> all lines. When we move
>> to the scheme of blocking read for re
In the read/write irq scheme, a program may open multiple irq devices. When
wrting (acking), we
need to find the user_intr_t that both points to the device port and has the
correct interrupt.
Thus the search_intr function need to match the interrupt id. For the
device_intr_ack method,
the id p
The tot_num_intr field is a count of how many deliverable interrupts across all
lines. When we move
to the scheme of blocking read for request and write for acking, it is possible
that an interrupt
can happen during a small period that the interrupt is acked, but the read has
not happended yet.
linux irq: remove suer_intr from linux_action
---
linux/dev/arch/i386/kernel/irq.c | 10 --
1 file changed, 10 deletions(-)
diff --git a/linux/dev/arch/i386/kernel/irq.c b/linux/dev/arch/i386/kernel/irq.c
index aee10462..de7d6c6a 100644
--- a/linux/dev/arch/i386/kernel/irq.c
+++ b/linux/
irqdev: make deliver_user_intr a linux irq handler
---
device/intr.c| 30 ++-
device/intr.h| 1 -
linux/dev/arch/i386/kernel/irq.c | 51 +---
3 files changed, 24 insertions(+), 58 deletions(-)
diff --git a/devic
This sets up the stage for making deliver_user_intr a linux irq handler, which
does not have a
return value. So we make deliver_user_intr return void, and move the check for
dead notification
port into intr_thread. Now deliver_user_intr does not touch main_intr_queue,
and only the server
thread
1. make deliver_intr a linux irq handler, and do not let it touch the
main_intr_queue. the queue
handling is solely in intr_thread (removing dead entries) and
insert_intr_entry (insertion).
2. remove user_intr from struct linux_action and its handling on linux side.
---
device/intr.c
1. The queue manipulated by two threads, the registration server and the
intr_thread, so we need a lock to protect it.
2. To initialize the lock, we introduce a irq_init function in
i386/i386/irq.[ch] and call it in device/device_init.c
---
device/device_init.c | 3 ++-
device/intr.c|
Ok. Will do.
Junling
> On Aug 2, 2020, at 4:52 PM, Samuel Thibault wrote:
>
> Junling Ma, le dim. 02 août 2020 16:45:22 -0700, a ecrit:
>> I guess you misunderstood what I mean. I mean that there seems to be no
>> interest in the discussion on IRC yesterday that we i
series of revised
patch for that.
Junling
> On Aug 2, 2020, at 4:33 PM, Samuel Thibault wrote:
>
> Junling Ma, le dim. 02 août 2020 16:19:39 -0700, a ecrit:
>> I was distracted by using a counter at all in the original code,
>> because an irq cannot be fired more than on
> On Aug 2, 2020, at 3:44 PM, Samuel Thibault wrote:
>
> Junling Ma, le dim. 02 août 2020 15:32:01 -0700, a ecrit:
>>>> 1. Change devicer_user_intr to be a linux interrupt handler.
>>>
>>> Right, this makes sense, but did you actually try it? Check the
Hi Samual,
Thanks for reviewing.
> Junling Ma, le sam. 01 août 2020 18:59:58 -0700, a ecrit:
>> PATCH 1 is included in this email: it prepares the stage by
>
> Please provide separate patches for each over these. Otherwise this is
> all conflated and way more difficult to r
In this patch, the old interface of device_intr_register and device_intr_ack is
removed.
Junling Ma
---
Makefrag.am| 2 -
device/ds_routines.c | 62 --
device/intr.c | 104 -
device/intr.h
Sorry the previous PATCH-2 was not in plain text. Here is a resend.
—
Hi all,
In this patch, the new interface of irq device is implemented. Please see the
first patch for a description.
Junling Ma
---
device/ds_routines.c | 2 +-
device/intr.c| 184
Hi all,
In this patch, the new interface of irq device is implemented. Please see the
first patch for a description.
Junling Ma
---
device/ds_routines.c | 2 +-
device/intr.c| 184 +--
device/intr.h| 11 ++-
i386/i386at/conf.c | 4
represent a specific interrupt line, so that notifications
queue on each line.
3. Make struct irqdev store a vector user_intr_t, indexed by interrupt line.
4. The n_unacked counter only increases when successfully delivered an
interrupt.
Any comments?
Best,
Junling Ma
---
device/device_init.c
21 matches
Mail list logo
