Skip to contentSkip to conversation

staging: fbtft: make dirty_lock IRQ-safe

PATCH In review Hui Su 3 participants 3 messages Aug 7 – Aug 8

Subsystems (3)
lore
Share & tools
lore

Hui Su

author
#0 lore raw

PATCH v2 1/1 staging: fbtft: make dirty_lock IRQ-safe

fbtft_mkdirty() can be reached from the fbcon rendering path while processing printk() in hardirq context. Meanwhile, dirty_lock is also taken by fbtft_deferred_io() in workqueue context with local interrupts enabled.

Lockdep reports a possible IRQ lock inversion involving dirty_lock and console_owner. A hardirq can interrupt a CPU holding dirty_lock and enter the console rendering path, which can attempt to acquire dirty_lock again.

The following lockdep report was observed on an RK3566 system with CONFIG_PROVE_LOCKING enabled:

WARNING: possible irq lock inversion dependency detected swapper/2/0 just changed the state of lock:
(console_owner){-...}-{0:0}
but this lock took another, HARDIRQ-unsafe lock in the past:
(&par->dirty_lock){+.+.}-{2:2}

  CPU0                    CPU1
  ----                    ----
  lock(&par->dirty_lock);
                         local_irq_disable();
                         lock(console_owner);
                         lock(&par->dirty_lock);
  <Interrupt>
    lock(console_owner);

*** DEADLOCK ***

Use spin_lock_irqsave() for fbtft_mkdirty() and spin_lock_irq() for fbtft_deferred_io(). They only access the dirty line range, so the IRQ-off regions remain short.

Fixes c296d5f9957c ("staging: fbtft: core support")

---
Changes in v2:
- Drop the comment explaining the lock ordering in fbtft_mkdirty().
- Use spin_lock_irq() in fbtft_deferred_io(), since it runs in workqueue context with interrupts enabled.

v1: https://lore.kernel.org/lkml/[email protected]/

 drivers/staging/fbtft/fbtft-core.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)
drivers/staging/fbtft/fbtft-core.c +5 -4
index ca0c38221c16..7925d974de80 100644
@@ -298,14 +298,15 @@ static void fbtft_mkdirty(struct fb_info *info, int y, int height)
 {
 	struct fbtft_par *par = info->par;
 	struct fb_deferred_io *fbdefio = info->fbdefio;
+	unsigned long flags;
 
 	/* Mark display lines/area as dirty */
-	spin_lock(&par->dirty_lock);
+	spin_lock_irqsave(&par->dirty_lock, flags);
 	if (y < par->dirty_lines_start)
 		par->dirty_lines_start = y;
 	if (y + height - 1 > par->dirty_lines_end)
 		par->dirty_lines_end = y + height - 1;
-	spin_unlock(&par->dirty_lock);
+	spin_unlock_irqrestore(&par->dirty_lock, flags);
 
 	/* Schedule deferred_io to update display (no-op if already on queue)*/
 	schedule_delayed_work(&info->deferred_work, fbdefio->delay);
@@ -318,13 +319,13 @@ static void fbtft_deferred_io(struct fb_info *info, struct list_head *pagereflis
 	struct fb_deferred_io_pageref *pageref;
 	unsigned int y_low = 0, y_high = 0;
 
-	spin_lock(&par->dirty_lock);
+	spin_lock_irq(&par->dirty_lock);
 	dirty_lines_start = par->dirty_lines_start;
 	dirty_lines_end = par->dirty_lines_end;
 	/* set display line markers as clean */
 	par->dirty_lines_start = par->info->var.yres - 1;
 	par->dirty_lines_end = 0;
-	spin_unlock(&par->dirty_lock);
+	spin_unlock_irq(&par->dirty_lock);
 
 	/* Mark display lines as dirty */
 	list_for_each_entry(pageref, pagereflist, list) {
Signature · 1 line
2.43.0
#1 lore raw
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] dirty_lock must be converted to raw_spinlock_t to prevent fatal kernel panics on PREEMPT_RT. --

--- Patch [1]: [PATCH v2] staging: fbtft: make dirty_lock IRQ-safe --- commit 2781cf2c7f0e0cfa744e4747b9f7004477fdff7b
Author: Hui Su <[email protected]>

staging: fbtft: make dirty_lock IRQ-safe

This commit updates the dirty_lock acquisitions in fbtft_mkdirty() and fbtft_deferred_io() to use IRQ-safe variants. This prevents a possible IRQ lock inversion between dirty_lock and console_owner.

drivers/staging/fbtft/fbtft-core.c static void fbtft_mkdirty(struct fb_info *info, int y, int height) · line 304

Quoted 13 lines
drivers/staging/fbtft/fbtft-core.c +2 -1
index ca0c38221c16b..7925d974de801 100644
@@ -298,14 +298,15 @@ static void fbtft_mkdirty(struct fb_info *info, int y, int height)
 {
 	struct fbtft_par *par = info->par;
 	struct fb_deferred_io *fbdefio = info->fbdefio;
+	unsigned long flags;
 
 	/* Mark display lines/area as dirty */
-	spin_lock(&par->dirty_lock);
+	spin_lock_irqsave(&par->dirty_lock, flags);

[Severity: High]
This isn't a bug introduced by this patch, but will this cause a kernel panic on PREEMPT_RT systems?

The commit message notes that fbtft_mkdirty() can be reached from the fbcon rendering path while processing printk() in hardirq context. On PREEMPT_RT, spinlock_t is mapped to a sleepable rt_mutex. Acquiring a sleepable lock in true hardirq context will trigger a sleep-in-atomic bug.

Since dirty_lock is accessed in true hardirq context, should the declaration of dirty_lock in struct fbtft_par be converted to a raw_spinlock_t and acquired via raw_spin_lock_irqsave() to safely synchronize under PREEMPT_RT?

Quoted 2 lines
drivers/staging/fbtft/fbtft-core.c +0 -0
 	if (y < par->dirty_lines_start)
 		par->dirty_lines_start = y;
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Reviewed-by on v2 1/1
#2 lore raw
In reply to Hui Su
Quoted 4 lines · Hui Su

fbtft_mkdirty() can be reached from the fbcon rendering path while processing printk() in hardirq context. Meanwhile, dirty_lock is also taken by fbtft_deferred_io() in workqueue context with local interrupts enabled.

...

Quoted 6 lines

Use spin_lock_irqsave() for fbtft_mkdirty() and spin_lock_irq() for fbtft_deferred_io(). They only access the dirty line range, so the IRQ-off regions remain short.

Fixes c296d5f9957c ("staging: fbtft: core support")