From: Gregory CLEMENT <gregory.clement@bootlin.com>
To: sashiko-reviews@lists.linux.dev
Cc: devicetree@vger.kernel.org, robh@kernel.org, conor+dt@kernel.org
Subject: Re: [PATCH v4 2/5] irqchip: irq-eyeq5-msi: add EyeQ5 PCIe MSI controller driver
Date: Fri, 20 Nov 2026 17:24:07 +0100 [thread overview]
Message-ID: <87a4n3pj2w.fsf@BLaptop.bootlin.com> (raw)
In-Reply-To: <sashiko-outbox-217188@kernel.org>
Hello,
> Thank you for your contribution! Sashiko AI review found 1 potential
> issue to consider:
> - [Medium] irqchip/eyeq5-msi: lost and unmasked interrupts in
> eyeq_msi_isr()
> --
>
> commit ecbaf749e93a909382eba764dafb43426071620b
> Author: Gregory CLEMENT <gregory.clement@bootlin.com>
>
> irqchip: irq-eyeq5-msi: add EyeQ5 PCIe MSI controller driver
>
> This commit adds an irqchip driver for the Mobileye EyeQ5 PCIe MSI
> controller, supporting both root complex PCI MSI and endpoint platform MSI
> roles.
>
>> diff --git a/drivers/irqchip/irq-eyeq5-msi.c b/drivers/irqchip/irq-eyeq5-msi.c
>> new file mode 100644
>> index 0000000000000..823dfb0b00b5b
>> --- /dev/null
>> +++ b/drivers/irqchip/irq-eyeq5-msi.c
> [ ... ]
>> +static struct irq_chip eyeq_msi_irq_chip = {
>> + .name = "EyeQ5-MSI",
>> + .irq_mask = eyeq_msi_mask_irq,
>> + .irq_unmask = eyeq_msi_unmask_irq,
>> + .irq_ack = eyeq_msi_ack_irq,
>> + .irq_compose_msi_msg = eyeq_msi_compose_msg,
>> +#ifdef CONFIG_SMP
>> + .irq_set_affinity = irq_chip_redirect_set_affinity,
>> +#endif
>> +};
>> +
>> +static void eyeq_msi_isr(struct irq_desc *desc)
>> +{
>> + struct irq_chip *parent_chip = irq_desc_get_chip(desc);
>> + struct eyeq_msi *msi = irq_desc_get_handler_data(desc);
>> + void __iomem *status_reg;
>> + unsigned long pending;
>> + unsigned int bit;
>> +
>> + guard_chained_irq(parent_chip, desc);
>> +
>> + status_reg = msi->gcsr_base + GCSR_IRQSTATUS;
>> +
>> + while (true) {
>> + pending = FIELD_GET(GCSR_IRQSTATUS_MASK, readw(status_reg));
>> + if (!pending)
>> + break;
>> +
>> + /*
>> + * Write-one-to-clear status register.
>> + */
>> + writew(FIELD_PREP(GCSR_IRQSTATUS_MASK, pending), status_reg);
>> +
>> + for_each_set_bit(bit, &pending, EYEQ_MSI_MAX_VECTORS) {
>> + if (!generic_handle_demux_domain_irq(msi->msi_domain, bit))
>> + pr_err_ratelimited("unmapped MSI interrupt %u\n", bit);
>> + }
>> + }
>> +}
>
> [Severity: Medium]
> Can eyeq_msi_isr() lose interrupts that arrive while a vector is lazily
> disabled, and dispatch vectors that are masked in GCSR_IRQMASK?
>
> eyeq_msi_domain_alloc() sets handle_edge_irq without IRQ_DISABLE_UNLAZY or
> an .irq_disable callback. When an MSI arrives after disable_irq(),
> eyeq_msi_isr() clears its bit in GCSR_IRQSTATUS before handle_edge_irq()
> records IRQS_PENDING and masks GCSR_IRQMASK. Because eyeq_msi_irq_chip
> has no .irq_retrigger callback and neither EYEQ5_MSI nor MIPS selects
> HARDIRQS_SW_RESEND, enable_irq() -> check_irq_resend() cannot replay
> IRQS_PENDING and the interrupt is lost.
>
> In addition, eyeq_msi_isr() does not mask pending with the enabled bits in
> GCSR_IRQMASK before clearing GCSR_IRQSTATUS and calling
> generic_handle_demux_domain_irq(). If a masked vector latches in
> GCSR_IRQSTATUS, a sibling vector firing will clear and dispatch the masked
> vector (for example re-entering a masked IRQF_ONESHOT threaded handler,
> since handle_edge_irq() checks irq_can_handle() rather than
> irqd_irq_masked()), or repeated writes to a masked vector can keep the
> while (true) loop spinning.
>
I reused my MSI test setup and added a new test: on the host, a helper
disables one MSI-X vector, raises that MSI-X from the endpoint, then
enables the vector again and checks its count in /proc/interrupts. This
confirmed the lost-interrupt issue was real and let me write the fix.
Since there was no further comment, I will send a v5 with the fix soon.
Grégory
> --
> Sashiko AI review · https://sashiko.dev/#/patchset/20261116-eyeq5_msi-v4-0-2e1233b014b0@bootlin.com?part=2
--
Grégory CLEMENT, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
next prev parent reply other threads:[~2026-11-20 16:24 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-16 13:58 [PATCH v4 0/5] irqchip: add EyeQ5 PCIe MSI controller Gregory CLEMENT
2026-11-16 13:58 ` [PATCH v4 1/5] dt-bindings: interrupt-controller: add Mobileye " Gregory CLEMENT
2026-11-16 13:58 ` [PATCH v4 2/5] irqchip: irq-eyeq5-msi: add EyeQ5 PCIe MSI controller driver Gregory CLEMENT
2026-11-16 14:43 ` sashiko-bot
2026-11-20 16:24 ` Gregory CLEMENT [this message]
2026-11-16 13:58 ` [PATCH v4 3/5] MIPS: eyeq: select the EyeQ5 MSI controller Gregory CLEMENT
2026-11-16 13:58 ` [PATCH v4 4/5] MIPS: dts: eyeq5: add MSI controllers Gregory CLEMENT
2026-11-16 13:58 ` [PATCH v4 5/5] MAINTAINERS: add EyeQ5 MSI controller entry Gregory CLEMENT
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=87a4n3pj2w.fsf@BLaptop.bootlin.com \
--to=gregory.clement@bootlin.com \
--cc=conor+dt@kernel.org \
--cc=devicetree@vger.kernel.org \
--cc=robh@kernel.org \
--cc=sashiko-reviews@lists.linux.dev \
/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