tty: moxa: unwind tty driver if PCI registration fails
Subsystems (2)
PATCH 1/2 tty: moxa: unwind tty driver if PCI registration fails
moxa_init() registers the tty driver before registering the PCI driver. If pci_register_driver() fails, module initialization returns without unregistering the tty driver or dropping its reference. moxa_exit() is not called after a failed module initialization.
Unregister the tty driver and drop its reference before returning the PCI registration error.
This issue was identified during our ongoing static-analysis research while reviewing kernel code.
--- drivers/tty/moxa.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/tty/moxa.c +4 -1
@@ -1172,8 +1172,11 @@ static int __init moxa_init(void) } retval =pci_register_driver (&moxa_pci_driver); -if (retval) +if (retval) {printk (KERN_ERR "Can't register MOXA pci driver!\n" ); +tty_unregister_driver (moxaDriver); +tty_driver_kref_put (moxaDriver); + }return retval; }
Signature · 1 line
2.47.1
PATCH 2/2 tty: moxa: use pcim_enable_device()
moxa_pci_probe() enables the PCI device with pci_enable_device(), but its failure paths and remove callback do not disable it. Repeated bind and unbind cycles can therefore leave the PCI enable count unbalanced.
Use pcim_enable_device() so the PCI core automatically disables the device when probe fails or the driver detaches. Keep the existing explicit BAR request and mapping cleanup unchanged.
This issue was identified during our ongoing static-analysis research while reviewing kernel code.
--- drivers/tty/moxa.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/tty/moxa.c +1 -1
@@ -1063,7 +1063,7 @@ static int moxa_pci_probe(struct pci_dev *pdev,int board_type = ent->driver_data;int retval; - retval =pci_enable_device (pdev); + retval =pcim_enable_device (pdev);if (retval) {dev_err (&pdev->dev,"can't enable pci device\n" ); goto err;
Signature · 1 line
2.47.1
Jiri Slaby
maintainerQuoted 7 lines · Myeonghun Pak
moxa_init()registers the tty driver before registering the PCI driver. Ifpci_register_driver()fails, module initialization returns without unregistering the tty driver or dropping its reference.moxa_exit()is not called after a failed module initialization.Unregister the tty driver and drop its reference before returning the PCI registration error.
I think, we should remove the whole driver instead:
https://lore.kernel.org/all/[email protected]/
Quoted 21 lines
--- drivers/tty/moxa.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)drivers/tty/moxa.c +4
-1@@ -1172,8 +1172,11 @@ static int __init moxa_init(void) } retval =pci_register_driver (&moxa_pci_driver); -if (retval) +if (retval) {printk (KERN_ERR "Can't register MOXA pci driver!\n" ); +tty_unregister_driver (moxaDriver); +tty_driver_kref_put (moxaDriver); + }return retval; }
Signature · 2 lines
js suse labs
Greg Kroah-Hartman
maintainerQuoted 10 lines · Jiri Slaby
Quoted 7 lines · Myeonghun Pak
moxa_init()registers the tty driver before registering the PCI driver. Ifpci_register_driver()fails, module initialization returns without unregistering the tty driver or dropping its reference.moxa_exit()is not called after a failed module initialization.Unregister the tty driver and drop its reference before returning the PCI registration error.
I think, we should remove the whole driver instead:
https://lore.kernel.org/all/[email protected]/
I agree, I'll submit a patch to just drop it.
thanks,
greg k-h
Hi Jiri and Greg,
Thanks for the review. I agree that removing the whole MOXA driver is the better direction, which supersedes this patch. Please drop/ignore this patch.
Best regards,
Myeonghun
2026년 8월 3일 (월) 오후 9:41, Greg Kroah-Hartman <[email protected]>님이 작성:
Quoted 18 lines · Jiri Slaby
Quoted 10 lines · Myeonghun Pak
Quoted 7 lines
moxa_init()registers the tty driver before registering the PCI driver. Ifpci_register_driver()fails, module initialization returns without unregistering the tty driver or dropping its reference.moxa_exit()is not called after a failed module initialization.Unregister the tty driver and drop its reference before returning the PCI registration error.
I think, we should remove the whole driver instead:
https://lore.kernel.org/all/[email protected]/I agree, I'll submit a patch to just drop it.
thanks,
greg k-h