i2c: sh_mobile: request IRQs after mapping registers
Subsystems (2)
PATCH 1/1 i2c: sh_mobile: request IRQs after mapping registers
The managed IRQ handlers are acquired before the managed register mapping, so devres unmaps the registers before freeing the IRQs. A transfer wakes its waiter before the ISR's final register read, and unregistering the adapter does not synchronize these IRQs. An in-flight handler can therefore still access the mapping when it is released during driver detach.
With forced-threaded IRQs, the handler can be preempted after waking the transfer waiter and before its final ICSR read.
Map the registers before requesting the IRQs so that managed cleanup frees and synchronizes all handlers before unmapping their registers. Keep IRQs available while unregistering the adapter because client removal may perform transfers.
This issue was identified as part of our ongoing research while reviewing kernel code.
---
Compile-tested with Clang, W=1. No hardware runtime testing.
Based on mainline 0fcb05b72344f40806cbe810de8872f1698bf053.
drivers/i2c/busses/i2c-sh_mobile.c +4 -4
@@ -880,10 +880,6 @@ static int sh_mobile_i2c_probe(struct platform_device *dev)return PTR_ERR (pd->clk); } - ret =sh_mobile_i2c_hook_irqs (dev, pd); -if (ret) -return ret; - pd->dev = &dev->dev;platform_set_drvdata (dev, pd);
@@ -891,6 +887,10 @@ static int sh_mobile_i2c_probe(struct platform_device *dev)if (IS_ERR (pd->reg))return PTR_ERR (pd->reg); + ret =sh_mobile_i2c_hook_irqs (dev, pd); +if (ret) +return ret; + ret =of_property_read_u32 (dev->dev.of_node,"clock-frequency" , &bus_speed); pd->bus_speed = (ret || !bus_speed) ?I2C_MAX_STANDARD_MODE_FREQ : bus_speed; pd->clks_per_count =1 ;
Andi Shyti
maintainerHi Myeounghun,
Quoted 15 lines · Myeonghun Pak
The managed IRQ handlers are acquired before the managed register mapping, so devres unmaps the registers before freeing the IRQs. A transfer wakes its waiter before the ISR's final register read, and unregistering the adapter does not synchronize these IRQs. An in-flight handler can therefore still access the mapping when it is released during driver detach.
With forced-threaded IRQs, the handler can be preempted after waking the transfer waiter and before its final ICSR read.
Map the registers before requesting the IRQs so that managed cleanup frees and synchronizes all handlers before unmapping their registers. Keep IRQs available while unregistering the adapter because client removal may perform transfers.
Quoted 4 lines
This issue was identified as part of our ongoing research while reviewing kernel code.
Fixes 7fe8a9993337 ("i2c: sh_mobile: devm conversion, irq setup")
Normally, I don't like taking fixes that don't address real issues, especially speculative AI-generated fixes.
However, this looks like a harmless reordering that seems correct, although the issue it addresses is rather hypothetical.
I'll remove the Fixes: tag, since this is purely speculative, and apply the patch to i2c/i2c.
Thanks,
Andi
Geert Uytterhoeven
Tested-byQuoted 24 lines · Andi Shyti
Quoted 15 lines · Myeonghun Pak
The managed IRQ handlers are acquired before the managed register mapping, so devres unmaps the registers before freeing the IRQs. A transfer wakes its waiter before the ISR's final register read, and unregistering the adapter does not synchronize these IRQs. An in-flight handler can therefore still access the mapping when it is released during driver detach.
With forced-threaded IRQs, the handler can be preempted after waking the transfer waiter and before its final ICSR read.
Map the registers before requesting the IRQs so that managed cleanup frees and synchronizes all handlers before unmapping their registers. Keep IRQs available while unregistering the adapter because client removal may perform transfers.
Quoted 4 lines
This issue was identified as part of our ongoing research while reviewing kernel code.
Fixes 7fe8a9993337 ("i2c: sh_mobile: devm conversion, irq setup")Normally, I don't like taking fixes that don't address real issues, especially speculative AI-generated fixes.
Quoted 6 lines
However, this looks like a harmless reordering that seems correct, although the issue it addresses is rather hypothetical.
I'll remove the Fixes: tag, since this is purely speculative, and apply the patch to i2c/i2c.
FWIW, no issues seen when boot-testing several platforms using this driver.
Gr{oetje,eeting}s,
Geert
Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected] In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
Quoted 15 lines · Myeonghun Pak
The managed IRQ handlers are acquired before the managed register mapping, so devres unmaps the registers before freeing the IRQs. A transfer wakes its waiter before the ISR's final register read, and unregistering the adapter does not synchronize these IRQs. An in-flight handler can therefore still access the mapping when it is released during driver detach.
With forced-threaded IRQs, the handler can be preempted after waking the transfer waiter and before its final ICSR read.
Map the registers before requesting the IRQs so that managed cleanup frees and synchronizes all handlers before unmapping their registers. Keep IRQs available while unregistering the adapter because client removal may perform transfers.
Quoted 7 lines
This issue was identified as part of our ongoing research while reviewing kernel code.
Fixes 7fe8a9993337 ("i2c: sh_mobile: devm conversion, irq setup")Co-developed-by Ijae Kim <[email protected]>Signed-off-by Ijae Kim <[email protected]>Signed-off-by Myeonghun Pak <[email protected]>
I know I am late to the game, but I still want to say that while changing the order to be "regs then irqs" is in general a good thing, it won't matter here, as I see it. Unless the later pm_runtime_get_sync() the device should not(tm) be able to send irqs. Famous last words...
Wolfram Sang
maintainerQuoted 3 lines
won't matter here, as I see it. Unless the later
pm_runtime_get_sync()the device should not(tm) be able to send irqs. Famous last words...
Which only matters during probe(), not during teardown... sorry for the noise! Patch is still good.