ipack: ipoctal: fix races and UAFs during module removal
cover ipack: ipoctal: fix races and UAFs during module removal
When the ipoctal device is removed while a userspace process still holds a tty fd open, several races and use-after-free bugs can be triggered. This series addresses the following issues:
1. UAF of struct ipoctal: the remove callback freed ipoctal via kfree() while in-flight tty ops could still be accessing it.
2. NULL dereference in ipoctal_write_tty(): __ipoctal_remove() freed xmit_buf while a concurrent write() could still dereference it.
3. UAF in ipoctal_cleanup(): ipack_put_carrier(ipoctal->dev) walked through ipoctal->dev to reach the carrier module, but the ipack_device could have already been freed by then.
4. Page faults due to devm_ioremap() regions being unmapped while tty ops still access hardware registers.
5. TOCTOU race between the removed-flag check and the subsequent hardware access in each tty op.
Patch 1 introduces kref-based lifetime management, caches the carrier owner module pointer to avoid chasing a dangling dev->bus pointer, adds a "removed" flag with checks in all tty ops that touch hardware, and adds a NULL guard on xmit_buf in the write path.
Patch 2 adds a read-write semaphore (remove_sem) to close the TOCTOU window: the tty ops hold the read lock for the duration of the hardware access, while __ipoctal_remove() acquires the write lock when setting the removed flag, ensuring that once removed is true no in-flight tty op is still inside a critical section.
Pei Xiao (2):
ipack: ipoctal: fix UAF, null-ptr-deref, and use-after-free in cleanup
on remove
ipack: ipoctal: add rwsem to guard against TOCTOU in remove path
drivers/ipack/devices/ipoctal.c | 60 ++++++++++++++++++++++++++++++--- 1 file changed, 56 insertions(+), 4 deletions(-)
--
2.25.1
PATCH 1/2 ipack: ipoctal: fix UAF, null-ptr-deref, and use-after-free in cleanup on remove
Three issues arise when the device is removed while a tty session is still active:
1. UAF of struct ipoctal: the remove callback frees ipoctal via kfree() while tty ops may still access it. Fix by introducing kref-based lifetime management — kref is taken in install() when a tty is opened and released in cleanup() when the tty is finally destroyed; remove() uses kref_put() instead of kfree().
2. NULL dereference in ipoctal_write_tty(): __ipoctal_remove() frees xmit_buf via tty_port_free_xmit_buf() while a userspace process may still hold the tty fd and call write(). Fix by checking for NULL xmit_buf in ipoctal_write_tty().
3. UAF in ipoctal_cleanup(): ipack_put_carrier(ipoctal->dev) dereferences ipoctal->dev after the ipack_device has been freed by ipack_device_del(). Fix by caching ipoctal->carrier_owner during probe() and calling module_put() on the cached pointer directly in cleanup(), avoiding any access to ipoctal->dev.
Also introduce a "removed" flag in struct ipoctal, set at the start of __ipoctal_remove(), and checked in every tty op that accesses hardware resources (port_activate, write_tty, set_termios, hangup, shutdown). This prevents page faults when devm_ioremap() regions are unmapped after remove() returns.
--- drivers/ipack/devices/ipoctal.c | 56 ++++++++++++++++++++++++++++++--- 1 file changed, 52 insertions(+), 4 deletions(-)
drivers/ipack/devices/ipoctal.c +52 -4
@@ -10,6 +10,7 @@#include <linux/device.h> #include <linux/module.h> #include <linux/interrupt.h> +#include <linux/kref.h> #include <linux/sched.h> #include <linux/tty.h> #include <linux/serial.h>
@@ -25,6 +26,8 @@static const struct tty_operations ipoctal_fops; +static void ipoctal_release (struct kref *kref); +struct ipoctal_channel {struct ipoctal_stats stats;unsigned int nb_bytes;
@@ -49,6 +52,9 @@ struct ipoctal {struct tty_driver *tty_drv;u8 __iomem *mem8_space;u8 __iomem *int_space; +struct kref kref; +struct module *carrier_owner; +bool removed; }; static inline struct ipoctal *chan_to_ipoctal(struct ipoctal_channel *chan,
@@ -70,8 +76,14 @@ static void ipoctal_reset_channel(struct ipoctal_channel *channel)static int ipoctal_port_activate (struct tty_port *port,struct tty_struct *tty) {struct ipoctal_channel *channel; +struct ipoctal *ipoctal; channel =dev_get_drvdata (tty->dev); + ipoctal =chan_to_ipoctal (channel, tty->index); + + +if (ipoctal->removed) +return -ENODEV ;/* * Enable RX. TX will be enabled when
@@ -95,6 +107,7 @@ static int ipoctal_install(struct tty_driver *driver, struct tty_struct *tty)if (res) goto err_put_carrier; +kref_get (&ipoctal->kref); tty->driver_data = channel;return 0 ;
@@ -460,8 +473,13 @@ static ssize_t ipoctal_write_tty(struct tty_struct *tty, const u8 *buf, size_t count) {struct ipoctal_channel *channel = tty->driver_data; +struct ipoctal *ipoctal =chan_to_ipoctal (channel, tty->index);size_t char_copied; + +if (ipoctal->removed || !channel->tty_port.xmit_buf) +return 0 ; + char_copied =ipoctal_copy_write_buffer (channel, buf, count);/* As the IP-OCTAL 485 only supports half duplex, do it manually */
@@ -501,8 +519,13 @@ static void ipoctal_set_termios(struct tty_struct *tty,unsigned char mr2 =0 ;unsigned char csr =0 ;struct ipoctal_channel *channel = tty->driver_data; +struct ipoctal *ipoctal =chan_to_ipoctal (channel, tty->index);speed_t baud; + +if (ipoctal->removed) +return ; + cflag = tty->termios.c_cflag;/* Disable and reset everything before change the setup */
@@ -631,10 +654,16 @@ static void ipoctal_hangup(struct tty_struct *tty) {unsigned long flags;struct ipoctal_channel *channel = tty->driver_data; +struct ipoctal *ipoctal;if (channel ==NULL )return ; + ipoctal =chan_to_ipoctal (channel, tty->index); + +if (ipoctal->removed) +return ; +spin_lock_irqsave (&channel->lock, flags); channel->nb_bytes =0 ; channel->pointer_read =0 ;
@@ -651,10 +680,16 @@ static void ipoctal_hangup(struct tty_struct *tty)static void ipoctal_shutdown (struct tty_struct *tty) {struct ipoctal_channel *channel = tty->driver_data; +struct ipoctal *ipoctal;if (channel ==NULL )return ; + ipoctal =chan_to_ipoctal (channel, tty->index); + +if (ipoctal->removed) +return ; +ipoctal_reset_channel (channel);tty_port_set_initialized (&channel->tty_port, false); }
@@ -664,8 +699,9 @@ static void ipoctal_cleanup(struct tty_struct *tty)struct ipoctal_channel *channel = tty->driver_data;struct ipoctal *ipoctal =chan_to_ipoctal (channel, tty->index); -/* release the carrier driver */ -ipack_put_carrier (ipoctal->dev); +/* release the carrier driver via cached owner */ +module_put (ipoctal->carrier_owner); +kref_put (&ipoctal->kref, ipoctal_release); }static const struct tty_operations ipoctal_fops = {
@@ -683,6 +719,13 @@ static const struct tty_operations ipoctal_fops = { .cleanup = ipoctal_cleanup, }; +static void ipoctal_release (struct kref *kref) +{ + struct ipoctal *ipoctal = container_of(kref, struct ipoctal, kref); + +kfree (ipoctal); +} +static int ipoctal_probe (struct ipack_device *dev) {int res;
@@ -692,7 +735,10 @@ static int ipoctal_probe(struct ipack_device *dev)if (ipoctal ==NULL )return -ENOMEM ; +kref_init (&ipoctal->kref); + ipoctal->dev = dev; + ipoctal->carrier_owner = dev->bus->owner; res =ipoctal_inst_slot (ipoctal, dev->bus->bus_nr, dev->slot);if (res) goto out_uninst;
@@ -701,7 +747,7 @@ static int ipoctal_probe(struct ipack_device *dev)return 0 ; out_uninst: -kfree (ipoctal); +kref_put (&ipoctal->kref, ipoctal_release);return res; }
@@ -709,6 +755,8 @@ static void __ipoctal_remove(struct ipoctal *ipoctal) {int i; + ipoctal->removed = true; + ipoctal->dev->bus->ops->free_irq (ipoctal->dev);for (i =0 ; i <NR_CHANNELS ; i++) {
@@ -725,7 +773,7 @@ static void __ipoctal_remove(struct ipoctal *ipoctal)tty_unregister_driver (ipoctal->tty_drv);kfree (ipoctal->tty_drv->name);tty_driver_kref_put (ipoctal->tty_drv); -kfree (ipoctal); +kref_put (&ipoctal->kref, ipoctal_release); }static void ipoctal_remove (struct ipack_device *idev)
Signature · 1 line
2.25.1
PATCH 2/2 ipack: ipoctal: add rwsem to guard against TOCTOU in remove path
In reply to Pei XiaoThe "removed" flag check in each tty op has a TOCTOU race with __ipoctal_remove(): the device could be removed between the flag check and the subsequent access to hardware resources (channel registers via iowrite8, or xmit_buf in write_tty).
Close this race by introducing a read-write semaphore (remove_sem). The tty ops acquire the read lock via guard(rwsem_read) for the full duration of the operation, while __ipoctal_remove() acquires the write lock via scoped_guard(rwsem_write) when setting the removed flag. This ensures that once removed is true, no in-flight tty op can still be accessing resources that are about to be freed by the remove path.
--- drivers/ipack/devices/ipoctal.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-)
drivers/ipack/devices/ipoctal.c +10 -6
@@ -11,6 +11,7 @@#include <linux/module.h> #include <linux/interrupt.h> #include <linux/kref.h> +#include <linux/rwsem.h> #include <linux/sched.h> #include <linux/tty.h> #include <linux/serial.h>
@@ -54,6 +55,7 @@ struct ipoctal {u8 __iomem *int_space;struct kref kref;struct module *carrier_owner; +struct rw_semaphore remove_sem;bool removed; };
@@ -81,7 +83,7 @@ static int ipoctal_port_activate(struct tty_port *port, struct tty_struct *tty) channel =dev_get_drvdata (tty->dev); ipoctal =chan_to_ipoctal (channel, tty->index); - +guard (rwsem_read)(&ipoctal->remove_sem);if (ipoctal->removed)return -ENODEV ;
@@ -476,7 +478,7 @@ static ssize_t ipoctal_write_tty(struct tty_struct *tty, const u8 *buf,struct ipoctal *ipoctal =chan_to_ipoctal (channel, tty->index);size_t char_copied; - +guard (rwsem_read)(&ipoctal->remove_sem);if (ipoctal->removed || !channel->tty_port.xmit_buf)return 0 ;
@@ -522,7 +524,7 @@ static void ipoctal_set_termios(struct tty_struct *tty,struct ipoctal *ipoctal =chan_to_ipoctal (channel, tty->index);speed_t baud; - +guard (rwsem_read)(&ipoctal->remove_sem);if (ipoctal->removed)return ;
@@ -660,7 +662,7 @@ static void ipoctal_hangup(struct tty_struct *tty)return ; ipoctal =chan_to_ipoctal (channel, tty->index); - +guard (rwsem_read)(&ipoctal->remove_sem);if (ipoctal->removed)return ;
@@ -686,7 +688,7 @@ static void ipoctal_shutdown(struct tty_struct *tty)return ; ipoctal =chan_to_ipoctal (channel, tty->index); - +guard (rwsem_read)(&ipoctal->remove_sem);if (ipoctal->removed)return ;
@@ -736,6 +738,7 @@ static int ipoctal_probe(struct ipack_device *dev)return -ENOMEM ;kref_init (&ipoctal->kref); +init_rwsem (&ipoctal->remove_sem); ipoctal->dev = dev; ipoctal->carrier_owner = dev->bus->owner;
@@ -755,7 +758,8 @@ static void __ipoctal_remove(struct ipoctal *ipoctal) {int i; - ipoctal->removed = true; +scoped_guard (rwsem_write, &ipoctal->remove_sem) + ipoctal->removed = true; ipoctal->dev->bus->ops->free_irq (ipoctal->dev);
Signature · 1 line
2.25.1
Greg KH
maintainerQuoted 3 lines · Pei Xiao
When the ipoctal device is removed while a userspace process still holds a tty fd open, several races and use-after-free bugs can be triggered. This series addresses the following issues:
Cool, but why remove the module at all?
And do you have this hardware to test with?
thanks,
greg k-h
Quoted 5 lines · Greg KH
Quoted 3 lines · Pei Xiao
When the ipoctal device is removed while a userspace process still holds a tty fd open, several races and use-after-free bugs can be triggered. This series addresses the following issues:
Cool, but why remove the module at all?
1. Carrier hot-unplug: the IPack carrier (e.g., TPCI-200) is a PCI device. When the carrier is physically removed or the PCI device is unbound via sysfs, ipack_bus_unregister() calls ipack_device_del() on every mezzanine device, which invokes ipoctal_remove(). This happens regardless of whether userspace still holds a tty fd open -- the carrier is gone, so the devices must go too. The ipack_get_carrier()/try_module_get() in ipoctal_install() only prevents rmmod of the carrier module, not physical removal.
2. Syzkaller/syzbot-style fuzzing: the original report from Shuangpeng Bai exercises exactly these paths. Even without real hardware, syzkaller can trigger the races by binding and unbinding drivers.
So while rmmod of ipoctal itself is blocked by the tty layer (via drv->owner), the remove callback can and will be called through carrier removal or driver unbind. The crashes are real.
Quoted 1 line
And do you have this hardware to test with?
I don't have the hardware. However:
- The fixes are structural and follow well-established kernel
patterns: kref for lifetime management, rwsem for TOCTOU races, and caching a module pointer to avoid chasing a freed bus pointer. These are the same idioms used across the tty layer and other subsystems.
The original reporter (Shuangpeng, CC'd) has tested this crash passed.
thanks for your responce and time!
Pei.
Thanks!
Quoted 3 lines
thanks,
greg k-h
Greg KH
maintainerQuoted 8 lines · Pei Xiao
Quoted 5 lines · Greg KH
Quoted 3 lines · Pei Xiao
When the ipoctal device is removed while a userspace process still holds a tty fd open, several races and use-after-free bugs can be triggered. This series addresses the following issues:
Cool, but why remove the module at all?
Quoted 9 lines
1. Carrier hot-unplug: the IPack carrier (e.g., TPCI-200) is a PCI device. When the carrier is physically removed or the PCI device is unbound via sysfs,
ipack_bus_unregister()callsipack_device_del()on every mezzanine device, which invokes ipoctal_remove(). This happens regardless of whether userspace still holds a tty fd open -- the carrier is gone, so the devices must go too. The ipack_get_carrier()/try_module_get() inipoctal_install()only prevents rmmod of the carrier module, not physical removal.
Yes, but does anyone actually remove the module or unbind it or remove it physically in a real system?
This is very very old hardware, removing it while active is not something that the hardware is designed to support, so why do we need to support it in the kernel?
Do you have this hardware?
Quoted 4 lines
2. Syzkaller/syzbot-style fuzzing: the original report from Shuangpeng Bai exercises exactly these paths. Even without real hardware, syzkaller can trigger the races by binding and unbinding drivers.
Sure, but remember bind/unbind is a DEBUG TOOL! And it is controlled by root, so it is a "best effort" thing that is there for kernel developers to use for working on their code. No real customer runs that code path (and if they do, they are an admin and can do much worse things to the system if they wanted to.)
Quoted 3 lines
So while rmmod of ipoctal itself is blocked by the tty layer (via
drv->owner), the remove callback can and will be called through carrier removal or driver unbind. The crashes are real.
But not a real use case ;)
Quoted 10 lines
Quoted 1 line
And do you have this hardware to test with?
I don't have the hardware. However:
- The fixes are structural and follow well-established kernel
patterns: kref for lifetime management, rwsem for TOCTOU races, and caching a module pointer to avoid chasing a freed bus pointer. These are the same idioms used across the tty layer and other subsystems.The original reporter (Shuangpeng, CC'd) has tested this crash passed.
With real hardware?
Ideally we could just remove the driver as I doubt anyone uses it anymore. You can't even buy the device.
thanks,
greg k-h
Quoted 50 lines · Greg KH
Quoted 7 lines · Pei Xiao
Quoted 5 lines · Greg KH
Quoted 3 lines · Pei Xiao
When the ipoctal device is removed while a userspace process still holds a tty fd open, several races and use-after-free bugs can be triggered. This series addresses the following issues:
Cool, but why remove the module at all?
Quoted 9 lines
1. Carrier hot-unplug: the IPack carrier (e.g., TPCI-200) is a PCI device. When the carrier is physically removed or the PCI device is unbound via sysfs,
ipack_bus_unregister()callsipack_device_del()on every mezzanine device, which invokes ipoctal_remove(). This happens regardless of whether userspace still holds a tty fd open -- the carrier is gone, so the devices must go too. The ipack_get_carrier()/try_module_get() inipoctal_install()only prevents rmmod of the carrier module, not physical removal.Yes, but does anyone actually remove the module or unbind it or remove it physically in a real system?
This is very very old hardware, removing it while active is not something that the hardware is designed to support, so why do we need to support it in the kernel?
Do you have this hardware?
Quoted 4 lines
2. Syzkaller/syzbot-style fuzzing: the original report from Shuangpeng Bai exercises exactly these paths. Even without real hardware, syzkaller can trigger the races by binding and unbinding drivers.
Sure, but remember bind/unbind is a DEBUG TOOL! And it is controlled by root, so it is a "best effort" thing that is there for kernel developers to use for working on their code. No real customer runs that code path (and if they do, they are an admin and can do much worse things to the system if they wanted to.)
Quoted 3 lines
So while rmmod of ipoctal itself is blocked by the tty layer (via
drv->owner), the remove callback can and will be called through carrier removal or driver unbind. The crashes are real.But not a real use case ;)
Quoted 9 lines
Quoted 1 line
And do you have this hardware to test with?
I don't have the hardware. However:
- The fixes are structural and follow well-established kernel
patterns: kref for lifetime management, rwsem for TOCTOU races, and caching a module pointer to avoid chasing a freed bus pointer. These are the same idioms used across the tty layer and other subsystems.
Quoted 5 lines
Quoted 1 line
The original reporter (Shuangpeng, CC'd) has tested this crash passed.
With real hardware?
Ideally we could just remove the driver as I doubt anyone uses it anymore. You can't even buy the device.
You are right, but I saw the phrase 'gladly review patches from others' and thought that although this is a very, very old driver, there might still be a
chance to get my changes merged. Haha, I really owe an apology to Shuangpeng, who tested my patches several times for me.
Also, I may have wasted your time, given how busy you are.
Pei.
Thanks!
Quoted 4 lines
thanks,
greg k-h