All of lore.kernel.org
 help / color / mirror / Atom feed
From: Lee Jones <lee@kernel.org>
To: lee@kernel.org, Thomas Bogendoerfer <tbogendoerfer@suse.de>,
	Stanislaw Skowronek <skylark@unaligned.org>,
	Joshua Kinard <linux@kumba.dev>,
	Brent Casavant <bcasavan@sgi.com>, Pat Gefre <pfg@sgi.com>,
	Paul Burton <paulburton@kernel.org>,
	mfd@lists.linux.dev, linux-kernel@vger.kernel.org
Cc: stable@vger.kernel.org
Subject: [PATCH v2 09/27] mfd: ioc3: Request parent IRQ via request_irq() and free before domain removal
Date: Tue, 27 Oct 2026 09:30:38 +0000	[thread overview]
Message-ID: <20261027093306.2812695-10-lee@kernel.org> (raw)
In-Reply-To: <20261027093306.2812695-1-lee@kernel.org>

ioc3_irq_domain_setup() registers ioc3_irq_handler() as a chained
interrupt handler using irq_set_chained_handler_and_data(), whilst
ioc3_mfd_remove() and the ioc3_mfd_probe() error path attempt to tear it
down with free_irq(ipd->domain_irq, (void *)ipd) after removing the IRQ
domain.  Because chained handlers bypass the standard irqaction list,
free_irq() warns and fails to detach ioc3_irq_handler(), leaving a
dangling handler referencing the freed IRQ domain and driver private
data.

Register ioc3_irq_handler() with request_irq() to match free_irq(),
dispatch child interrupts via generic_handle_domain_irq_safe(), and free
the IRQ prior to removing the IRQ domain on teardown.

Fixes: 0ce5ebd24d25 ("mfd: ioc3: Add driver for SGI IOC3 chip")
Closes: https://sashiko.dev/#/bug/linux-81f02420-548c-47cf-bd46-0bdcfb5c1738
Cc: stable@vger.kernel.org
Signed-off-by: Lee Jones <lee@kernel.org>
---
 drivers/mfd/ioc3.c | 32 ++++++++++++++++++++------------
 1 file changed, 20 insertions(+), 12 deletions(-)

diff --git a/drivers/mfd/ioc3.c b/drivers/mfd/ioc3.c
index 5f8ac364b610..885f19154e02 100644
--- a/drivers/mfd/ioc3.c
+++ b/drivers/mfd/ioc3.c
@@ -99,10 +99,9 @@ static const struct irq_domain_ops ioc3_irq_domain_ops = {
 	.unmap = ioc3_irq_domain_unmap,
 };
 
-static void ioc3_irq_handler(struct irq_desc *desc)
+static irqreturn_t ioc3_irq_handler(int irq, void *data)
 {
-	struct irq_domain *domain = irq_desc_get_handler_data(desc);
-	struct ioc3_priv_data *ipd = domain->host_data;
+	struct ioc3_priv_data *ipd = data;
 	struct ioc3 __iomem *regs = ipd->regs;
 	u32 pending, mask;
 
@@ -110,10 +109,12 @@ static void ioc3_irq_handler(struct irq_desc *desc)
 	mask = readl(&regs->sio_ies);
 	pending &= mask; /* Mask off not enabled interrupts */
 
-	if (pending)
-		generic_handle_domain_irq(domain, __ffs(pending));
-	else
-		spurious_interrupt();
+	if (pending) {
+		generic_handle_domain_irq_safe(ipd->domain, __ffs(pending));
+		return IRQ_HANDLED;
+	}
+
+	return IRQ_NONE;
 }
 
 /*
@@ -133,6 +134,7 @@ static int ioc3_irq_domain_setup(struct ioc3_priv_data *ipd, int irq)
 {
 	struct irq_domain *domain;
 	struct fwnode_handle *fn;
+	int ret = -ENOMEM;
 
 	fn = irq_domain_alloc_named_fwnode("IOC3");
 	if (!fn)
@@ -146,13 +148,19 @@ static int ioc3_irq_domain_setup(struct ioc3_priv_data *ipd, int irq)
 
 	ipd->domain = domain;
 
-	irq_set_chained_handler_and_data(irq, ioc3_irq_handler, domain);
+	ret = request_irq(irq, ioc3_irq_handler, IRQF_SHARED, "IOC3", ipd);
+	if (ret) {
+		irq_domain_remove(domain);
+		irq_domain_free_fwnode(fn);
+		ipd->domain = NULL;
+		goto err;
+	}
 	ipd->domain_irq = irq;
 	return 0;
 
 err:
 	dev_err(&ipd->pdev->dev, "irq domain setup failed\n");
-	return -ENOMEM;
+	return ret;
 }
 
 static const struct resource ioc3_uarta_resources[] = {
@@ -615,9 +623,9 @@ static int ioc3_mfd_probe(struct pci_dev *pdev,
 		if (ipd->domain) {
 			struct fwnode_handle *fn = ipd->domain->fwnode;
 
+			free_irq(ipd->domain_irq, (void *)ipd);
 			irq_domain_remove(ipd->domain);
 			irq_domain_free_fwnode(fn);
-			free_irq(ipd->domain_irq, (void *)ipd);
 		}
 		pci_iounmap(pdev, regs);
 		goto out_disable_device;
@@ -645,9 +653,9 @@ static void ioc3_mfd_remove(struct pci_dev *pdev)
 	if (ipd->domain) {
 		struct fwnode_handle *fn = ipd->domain->fwnode;
 
-		irq_domain_remove(ipd->domain);
-		irq_domain_free_fwnode(fn);
 		free_irq(ipd->domain_irq, (void *)ipd);
+		irq_domain_remove(ipd->domain);
+		irq_domain_free_fwnode(fn);
 	}
 	pci_iounmap(pdev, ipd->regs);
 	pci_disable_device(pdev);
-- 
2.56.0.307.g5fe7bd680c-goog


  parent reply	other threads:[~2026-10-27 10:37 UTC|newest]

Thread overview: 36+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-27  9:30 [PATCH v2 00/27] mfd: Assorted resource lifetime, teardown, and error-handling fixes Lee Jones
2026-10-27  9:30 ` [PATCH v2 01/27] mfd: nct6694: Allocate USB DMA buffers with kzalloc_obj() Lee Jones
2026-10-27  9:30 ` [PATCH v2 02/27] mfd: ls2kbmc: Unregister GPIO GSI via devm_add_action_or_reset() Lee Jones
2026-10-27  9:30 ` [PATCH v2 03/27] mfd: ls2kbmc: Cancel reset work on teardown via devm_work_autocancel() Lee Jones
2026-10-27  9:30 ` [PATCH v2 04/27] mfd: ls2kbmc: Allocate simplefb platform_data on the heap Lee Jones
2026-10-27  9:30 ` [PATCH v2 05/27] mfd: bd96801: Allocate watchdog IRQ resource " Lee Jones
2026-10-27  9:30 ` [PATCH v2 06/27] mfd: ti_am335x_tscadc: Retrieve parent driver data via dev_get_drvdata() Lee Jones
2026-10-27  9:30 ` [PATCH v2 07/27] mfd: db8500-prcmu: Make AB8500 cell and IRQ resource static Lee Jones
2026-10-27  9:30 ` [PATCH v2 08/27] mfd: nct6694: Do not resubmit interrupt URB on endpoint stall (-EPIPE) Lee Jones
2026-10-27  9:30 ` Lee Jones [this message]
2026-10-27  9:30 ` [PATCH v2 10/27] mfd: intel-lpss: Remove registered MFD child devices on probe failure Lee Jones
2026-10-27  9:30 ` [PATCH v2 11/27] mfd: max8907: Manage power-off handler via devm_register_sys_off_handler() Lee Jones
2026-10-27  9:30 ` [PATCH v2 12/27] mfd: twl-core: Clear pm_power_off callback in twl_remove() Lee Jones
2026-10-27  9:30 ` [PATCH v2 13/27] mfd: twl-core: Clear twl_priv pointer on teardown and probe failure Lee Jones
2026-10-28 15:50   ` Andreas Kemnade
2026-10-30 10:48     ` Lee Jones
2026-10-30 10:52       ` Andreas Kemnade
2026-10-30 11:28         ` Lee Jones
2026-10-30 12:10           ` Andreas Kemnade
2026-11-02 10:13             ` Lee Jones
2026-10-27  9:30 ` [PATCH v2 14/27] mfd: loongson-se: Check return value of dma_set_mask_and_coherent() Lee Jones
2026-10-27  9:30 ` [PATCH v2 15/27] mfd: loongson-se: Zero-initialise controller command structures Lee Jones
2026-10-27  9:30 ` [PATCH v2 16/27] mfd: loongson-se: Reinitialise completions and wait uninterruptibly for commands Lee Jones
2026-10-27  9:30 ` [PATCH v2 17/27] mfd: loongson-se: Zero-allocate driver data and initialise engine completions in probe Lee Jones
2026-10-27  9:30 ` [PATCH v2 18/27] mfd: loongson-se: Prevent out-of-bounds engine access in se_irq_handler() Lee Jones
2026-10-27  9:30 ` [PATCH v2 19/27] mfd: loongson-se: Stop controller and mask interrupts on teardown and shutdown Lee Jones
2026-10-27  9:30 ` [PATCH v2 20/27] mfd: loongson-se: Validate engine ID, DMA buffer size, and IRQ registration Lee Jones
2026-10-27  9:30 ` [PATCH v2 21/27] mfd: abx500-core: Unregister device ops on devm release Lee Jones
2026-10-27  9:30 ` [PATCH v2 22/27] mfd: ab8500-core: Register sysfs attribute groups via devm_device_add_group() Lee Jones
2026-10-27  9:30 ` [PATCH v2 23/27] mfd: tps65217: Manage IRQ domain teardown via devres Lee Jones
2026-10-27  9:30 ` [PATCH v2 24/27] mfd: mt6397: Manage IRQ domain lifecycle with devres Lee Jones
2026-10-28  3:45   ` Chen-Yu Tsai
2026-10-27  9:30 ` [PATCH v2 25/27] mfd: mt6397: Unregister PM notifier on probe failure and unbind Lee Jones
2026-10-28  4:15   ` Chen-Yu Tsai
2026-10-27  9:30 ` [PATCH v2 26/27] i2c: cgbc: Propagate hardware errors when polling read status Lee Jones
2026-10-27  9:30 ` [PATCH v2 27/27] mfd: ti_am335x_tscadc: Remove redundant ti_tscadc_dev_get() helper Lee Jones

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=20261027093306.2812695-10-lee@kernel.org \
    --to=lee@kernel.org \
    --cc=bcasavan@sgi.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux@kumba.dev \
    --cc=mfd@lists.linux.dev \
    --cc=paulburton@kernel.org \
    --cc=pfg@sgi.com \
    --cc=skylark@unaligned.org \
    --cc=stable@vger.kernel.org \
    --cc=tbogendoerfer@suse.de \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.