[PATCH 1/2] fb: avoid oops when fb is removed by remove_conflicting_framebuffers

Subsystems: framebuffer layer, the rest

DORMANTno replies

2 messages, 1 author, 2011-01-18 · open the first message on its own page

[PATCH 1/2] fb: avoid oops when fb is removed by remove_conflicting_framebuffers

From: Herton Ronaldo Krzesinski <hidden>
Date: 2011-01-18 23:08:23

Firmware framebuffers (which have the flag FBINFO_MISC_FIRMWARE) can be
removed and replaced by another native framebuffer, like what happens
when you boot with a vesa framebuffer and later loads a drm module with
modesetting enabled on x86.

When framebuffer which is going to replace the firmware framebuffer is
registered, unregister_framebuffer is called for the old firmware
framebuffer automatically inside remove_conflicting_framebuffers
function. The problem is that while this happens, the old framebuffer
can still be in use.

The first issue in this is that unregister_framebuffer can free/destroy
struct fb_info when calling fb_info->fbops->fb_destroy. The same struct
fb_info is given to file->private_data inside fb_open function, so if
there is an application that have old framebuffer open, and it closes it
after the framebuffer was replaced, fb_release will still get the old
struct fb_info from file->private_data. As it was freed, the kernel will
most likely oops accessing an already freed location inside fb_release.
To fix this, add a reference count to struct fb_info and use it, so the
struct will be only destroyed after the last user closes the framebuffer.

The second issue is that the file_operations framebuffer functions
reference registered_fb array without locking. This can cause a race and
oops if application using the old firmware framebuffer calls a
read/write/mmap/ioctl function while unregister_framebuffer is running,
in the window inside it where registered_fb for the old framebuffer is
set to NULL. To avoid this situation, do not get the framebuffer from
registered_fb array, instead get it from file->private_data which is set
on fb_open. While at it, also add a new fb_info state which is set when
the framebuffer is being unregistered, so that file_operations functions
can check and prevent access to framebuffer when possible, returning an
error if the framebuffer is being or already unregistered.

This addresses kernel.org bug #26232. A testcase is provided in the
ticket which triggers the problem on current kernels.

References: https://bugzilla.kernel.org/show_bug.cgi?id&232
Signed-off-by: Herton Ronaldo Krzesinski <redacted>
---
 drivers/video/fbmem.c   |  170 ++++++++++++++++++++++++++++++++++-------------
 drivers/video/fbsysfs.c |    2 +
 include/linux/fb.h      |    2 +
 3 files changed, 127 insertions(+), 47 deletions(-)
diff --git a/drivers/video/fbmem.c b/drivers/video/fbmem.c
index 4ac1201..05d09ac 100644
--- a/drivers/video/fbmem.c
+++ b/drivers/video/fbmem.c
@@ -48,7 +48,7 @@ int num_registered_fb __read_mostly;
 int lock_fb_info(struct fb_info *info)
 {
 	mutex_lock(&info->lock);
-	if (!info->fbops) {
+	if (unlikely(!info->fbops || info->state = FBINFO_STATE_EXITING)) {
 		mutex_unlock(&info->lock);
 		return 0;
 	}
@@ -56,6 +56,72 @@ int lock_fb_info(struct fb_info *info)
 }
 EXPORT_SYMBOL(lock_fb_info);
 
+static int fbops_lock_fb_info(struct fb_info *info)
+{
+	int err;
+
+	if (!info->screen_base)
+		return -ENODEV;
+
+	if (info->flags = FBINFO_MISC_FIRMWARE) {
+		err = lock_fb_info(info);
+		if (err) {
+			if (info->state = FBINFO_STATE_SUSPENDED)
+				err = -EPERM;
+			if (err < 1)
+				unlock_fb_info(info);
+			else
+				err = 0;
+		} else
+			err = -ENODEV;
+		return err;
+	}
+
+	if (info->state != FBINFO_STATE_RUNNING)
+		return -EPERM;
+
+	return 0;
+}
+
+static void fbops_unlock_fb_info(struct fb_info *info)
+{
+	if (info->flags = FBINFO_MISC_FIRMWARE)
+		unlock_fb_info(info);
+}
+
+static void fb_kref_init(struct fb_info *info)
+{
+	if (info->flags = FBINFO_MISC_FIRMWARE)
+		kref_init(&info->fwfb_rfcnt);
+}
+
+static void fb_kref_get(struct fb_info *info)
+{
+	if (info->flags = FBINFO_MISC_FIRMWARE)
+		kref_get(&info->fwfb_rfcnt);
+}
+
+static void fb_kref_release(struct kref *ref)
+{
+	struct fb_info *info = container_of(ref, struct fb_info, fwfb_rfcnt);
+
+	if (info->fbops->fb_destroy)
+		info->fbops->fb_destroy(info);
+}
+
+static int fb_kref_put(struct fb_info *info, bool in_unregister)
+{
+	if (info->flags = FBINFO_MISC_FIRMWARE)
+		return kref_put(&info->fwfb_rfcnt, fb_kref_release);
+
+	if (in_unregister && info->fbops->fb_destroy) {
+		info->fbops->fb_destroy(info);
+		return 1;
+	}
+
+	return 0;
+}
+
 /*
  * Helpers
  */
@@ -694,30 +760,30 @@ static ssize_t
 fb_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
 {
 	unsigned long p = *ppos;
-	struct inode *inode = file->f_path.dentry->d_inode;
-	int fbidx = iminor(inode);
-	struct fb_info *info = registered_fb[fbidx];
+	struct fb_info *info = file->private_data;
 	u8 *buffer, *dst;
 	u8 __iomem *src;
-	int c, cnt = 0, err = 0;
+	int c, cnt = 0, err;
 	unsigned long total_size;
 
-	if (!info || ! info->screen_base)
-		return -ENODEV;
+	err = fbops_lock_fb_info(info);
+	if (err)
+		return err;
 
-	if (info->state != FBINFO_STATE_RUNNING)
-		return -EPERM;
+	if (info->fbops->fb_read) {
+		err = info->fbops->fb_read(info, buf, count, ppos);
+		goto fb_read_exit;
+	}
 
-	if (info->fbops->fb_read)
-		return info->fbops->fb_read(info, buf, count, ppos);
-	
 	total_size = info->screen_size;
 
 	if (total_size = 0)
 		total_size = info->fix.smem_len;
 
-	if (p >= total_size)
-		return 0;
+	if (p >= total_size) {
+		err = 0;
+		goto fb_read_exit;
+	}
 
 	if (count >= total_size)
 		count = total_size;
@@ -727,8 +793,10 @@ fb_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
 
 	buffer = kmalloc((count > PAGE_SIZE) ? PAGE_SIZE : count,
 			 GFP_KERNEL);
-	if (!buffer)
-		return -ENOMEM;
+	if (!buffer) {
+		err = -ENOMEM;
+		goto fb_read_exit;
+	}
 
 	src = (u8 __iomem *) (info->screen_base + p);
 
@@ -754,37 +822,42 @@ fb_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
 
 	kfree(buffer);
 
-	return (err) ? err : cnt;
+	if (!err)
+		err = cnt;
+
+fb_read_exit:
+	fbops_unlock_fb_info(info);
+	return err;
 }
 
 static ssize_t
 fb_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
 {
 	unsigned long p = *ppos;
-	struct inode *inode = file->f_path.dentry->d_inode;
-	int fbidx = iminor(inode);
-	struct fb_info *info = registered_fb[fbidx];
+	struct fb_info *info = file->private_data;
 	u8 *buffer, *src;
 	u8 __iomem *dst;
-	int c, cnt = 0, err = 0;
+	int c, cnt = 0, err;
 	unsigned long total_size;
 
-	if (!info || !info->screen_base)
-		return -ENODEV;
+	err = fbops_lock_fb_info(info);
+	if (err)
+		return err;
 
-	if (info->state != FBINFO_STATE_RUNNING)
-		return -EPERM;
+	if (info->fbops->fb_write) {
+		err = info->fbops->fb_write(info, buf, count, ppos);
+		goto fb_write_exit;
+	}
 
-	if (info->fbops->fb_write)
-		return info->fbops->fb_write(info, buf, count, ppos);
-	
 	total_size = info->screen_size;
 
 	if (total_size = 0)
 		total_size = info->fix.smem_len;
 
-	if (p > total_size)
-		return -EFBIG;
+	if (p > total_size) {
+		err = -EFBIG;
+		goto fb_write_exit;
+	}
 
 	if (count > total_size) {
 		err = -EFBIG;
@@ -800,8 +873,10 @@ fb_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
 
 	buffer = kmalloc((count > PAGE_SIZE) ? PAGE_SIZE : count,
 			 GFP_KERNEL);
-	if (!buffer)
-		return -ENOMEM;
+	if (!buffer) {
+		err = -ENOMEM;
+		goto fb_write_exit;
+	}
 
 	dst = (u8 __iomem *) (info->screen_base + p);
 
@@ -828,7 +903,12 @@ fb_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
 
 	kfree(buffer);
 
-	return (cnt) ? cnt : err;
+	if (cnt)
+		err = cnt;
+
+fb_write_exit:
+	fbops_unlock_fb_info(info);
+	return err;
 }
 
 int
@@ -1141,9 +1221,7 @@ static long do_fb_ioctl(struct fb_info *info, unsigned int cmd,
 
 static long fb_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
-	struct inode *inode = file->f_path.dentry->d_inode;
-	int fbidx = iminor(inode);
-	struct fb_info *info = registered_fb[fbidx];
+	struct fb_info *info = file->private_data;
 
 	return do_fb_ioctl(info, cmd, arg);
 }
@@ -1265,9 +1343,7 @@ static int fb_get_fscreeninfo(struct fb_info *info, unsigned int cmd,
 static long fb_compat_ioctl(struct file *file, unsigned int cmd,
 			    unsigned long arg)
 {
-	struct inode *inode = file->f_path.dentry->d_inode;
-	int fbidx = iminor(inode);
-	struct fb_info *info = registered_fb[fbidx];
+	struct fb_info *info = file->private_data;
 	struct fb_ops *fb = info->fbops;
 	long ret = -ENOIOCTLCMD;
 
@@ -1303,8 +1379,7 @@ static long fb_compat_ioctl(struct file *file, unsigned int cmd,
 static int
 fb_mmap(struct file *file, struct vm_area_struct * vma)
 {
-	int fbidx = iminor(file->f_path.dentry->d_inode);
-	struct fb_info *info = registered_fb[fbidx];
+	struct fb_info *info = file->private_data;
 	struct fb_ops *fb = info->fbops;
 	unsigned long off;
 	unsigned long start;
@@ -1380,6 +1455,8 @@ __releases(&info->lock)
 		if (res)
 			module_put(info->fbops->owner);
 	}
+	if (!res)
+		fb_kref_get(info);
 #ifdef CONFIG_FB_DEFERRED_IO
 	if (info->fbdefio)
 		fb_deferred_io_open(info, inode, file);
@@ -1401,6 +1478,7 @@ __releases(&info->lock)
 		info->fbops->fb_release(info,1);
 	module_put(info->fbops->owner);
 	mutex_unlock(&info->lock);
+	fb_kref_put(info, false);
 	return 0;
 }
 
@@ -1549,6 +1627,7 @@ register_framebuffer(struct fb_info *fb_info)
 	fb_info->node = i;
 	mutex_init(&fb_info->lock);
 	mutex_init(&fb_info->mm_lock);
+	fb_kref_init(fb_info);
 
 	fb_info->dev = device_create(fb_class, fb_info->device,
 				     MKDEV(FB_MAJOR, i), NULL, "fb%d", i);
@@ -1622,9 +1701,9 @@ unregister_framebuffer(struct fb_info *fb_info)
 		goto done;
 	}
 
-
 	if (!lock_fb_info(fb_info))
 		return -ENODEV;
+	fb_info->state = FBINFO_STATE_EXITING;
 	event.info = fb_info;
 	ret = fb_notifier_call_chain(FB_EVENT_FB_UNBIND, &event);
 	unlock_fb_info(fb_info);
@@ -1644,10 +1723,7 @@ unregister_framebuffer(struct fb_info *fb_info)
 	device_destroy(fb_class, MKDEV(FB_MAJOR, i));
 	event.info = fb_info;
 	fb_notifier_call_chain(FB_EVENT_FB_UNREGISTERED, &event);
-
-	/* this may free fb info */
-	if (fb_info->fbops->fb_destroy)
-		fb_info->fbops->fb_destroy(fb_info);
+	fb_kref_put(fb_info, true);
 done:
 	return ret;
 }
@@ -1655,7 +1731,7 @@ done:
 /**
  *	fb_set_suspend - low level driver signals suspend
  *	@info: framebuffer affected
- *	@state: 0 = resuming, !=0 = suspending
+ *	@state: 0 = resuming, 1 = suspending
  *
  *	This is meant to be used by low level drivers to
  * 	signal suspend/resume to the core & clients.
diff --git a/drivers/video/fbsysfs.c b/drivers/video/fbsysfs.c
index 0a08f13..be361b5 100644
--- a/drivers/video/fbsysfs.c
+++ b/drivers/video/fbsysfs.c
@@ -398,6 +398,8 @@ static ssize_t store_fbstate(struct device *device,
 	char *last = NULL;
 
 	state = simple_strtoul(buf, &last, 0);
+	if (state && state > 1)
+		return -EINVAL;
 
 	acquire_console_sem();
 	fb_set_suspend(fb_info, (int)state);
diff --git a/include/linux/fb.h b/include/linux/fb.h
index d1631d3..836f605 100644
--- a/include/linux/fb.h
+++ b/include/linux/fb.h
@@ -871,6 +871,7 @@ struct fb_info {
 	void *pseudo_palette;		/* Fake palette of 16 colors */ 
 #define FBINFO_STATE_RUNNING	0
 #define FBINFO_STATE_SUSPENDED	1
+#define FBINFO_STATE_EXITING	2
 	u32 state;			/* Hardware state i.e suspend */
 	void *fbcon_par;                /* fbcon use-only private area */
 	/* From here on everything is device dependent */
@@ -885,6 +886,7 @@ struct fb_info {
 			resource_size_t size;
 		} ranges[0];
 	} *apertures;
+	struct kref fwfb_rfcnt;
 };
 
 static inline struct apertures_struct *alloc_apertures(unsigned int max_num) {
-- 
1.7.3.4

[PATCH 2/2] fb: avoid possible deadlock caused by fb_set_suspend

From: Herton Ronaldo Krzesinski <hidden>
Date: 2011-01-18 23:08:21

A lock ordering issue can cause deadlocks: in framebuffer/console code,
all needed struct fb_info locks are taken before acquire_console_sem(),
in places which need to take console semaphore.

But fb_set_suspend is always called with console semaphore held, and
inside it we call lock_fb_info which gets the fb_info lock, inverse
locking order of what the rest of the code does. This causes a real
deadlock issue, when we write to state fb sysfs attribute (which calls
fb_set_suspend) while a framebuffer is being unregistered by
remove_conflicting_framebuffers, as can be shown by following show
blocked state trace on a test program which loads i915 and runs another
forked processes writing to state attribute:

Test process with semaphore held and trying to get fb_info lock:
...
fb-test2      D 0000000000000000     0   237    228 0x00000000
 ffff8800774f3d68 0000000000000082 00000000000135c0 00000000000135c0
 ffff880000000000 ffff8800774f3fd8 ffff8800774f3fd8 ffff880076ee4530
 00000000000135c0 ffff8800774f3fd8 ffff8800774f2000 00000000000135c0
Call Trace:
 [<ffffffff8141287a>] __mutex_lock_slowpath+0x11a/0x1e0
 [<ffffffff814142f2>] ? _raw_spin_lock_irq+0x22/0x40
 [<ffffffff814123d3>] mutex_lock+0x23/0x50
 [<ffffffff8125dfc5>] lock_fb_info+0x25/0x60
 [<ffffffff8125e3f0>] fb_set_suspend+0x20/0x80
 [<ffffffff81263e2f>] store_fbstate+0x4f/0x70
 [<ffffffff812e7f70>] dev_attr_store+0x20/0x30
 [<ffffffff811c46b4>] sysfs_write_file+0xd4/0x160
 [<ffffffff81155a26>] vfs_write+0xc6/0x190
 [<ffffffff81155d51>] sys_write+0x51/0x90
 [<ffffffff8100c012>] system_call_fastpath+0x16/0x1b
...
modprobe process stalled because has the fb_info lock (got inside
unregister_framebuffer) but waiting for the semaphore held by the
test process which is waiting to get the fb_info lock:
...
modprobe      D 0000000000000000     0   230    218 0x00000000
 ffff880077a4d618 0000000000000082 0000000000000001 0000000000000001
 ffff880000000000 ffff880077a4dfd8 ffff880077a4dfd8 ffff8800775a2e20
 00000000000135c0 ffff880077a4dfd8 ffff880077a4c000 00000000000135c0
Call Trace:
 [<ffffffff81411fe5>] schedule_timeout+0x215/0x310
 [<ffffffff81058051>] ? get_parent_ip+0x11/0x50
 [<ffffffff814130dd>] __down+0x6d/0xb0
 [<ffffffff81089f71>] down+0x41/0x50
 [<ffffffff810629ac>] acquire_console_sem+0x2c/0x50
 [<ffffffff812ca53d>] unbind_con_driver+0xad/0x2d0
 [<ffffffff8126f5f7>] fbcon_event_notify+0x457/0x890
 [<ffffffff814144ff>] ? _raw_spin_unlock_irqrestore+0x1f/0x50
 [<ffffffff81058051>] ? get_parent_ip+0x11/0x50
 [<ffffffff8141836d>] notifier_call_chain+0x4d/0x70
 [<ffffffff8108a3b8>] __blocking_notifier_call_chain+0x58/0x80
 [<ffffffff8108a3f6>] blocking_notifier_call_chain+0x16/0x20
 [<ffffffff8125dabb>] fb_notifier_call_chain+0x1b/0x20
 [<ffffffff8125e6ac>] unregister_framebuffer+0x7c/0x130
 [<ffffffff8125e8b3>] remove_conflicting_framebuffers+0x153/0x180
 [<ffffffff8125eef3>] register_framebuffer+0x93/0x2c0
 [<ffffffffa0331112>] drm_fb_helper_single_fb_probe+0x252/0x2f0 [drm_kms_helper]
 [<ffffffffa03314a3>] drm_fb_helper_initial_config+0x2f3/0x6d0 [drm_kms_helper]
 [<ffffffffa03318dd>] ? drm_fb_helper_single_add_all_connectors+0x5d/0x1c0 [drm_kms_helper]
 [<ffffffffa037b588>] intel_fbdev_init+0xa8/0x160 [i915]
 [<ffffffffa0343d74>] i915_driver_load+0x854/0x12b0 [i915]
 [<ffffffffa02f0e7e>] drm_get_pci_dev+0x19e/0x360 [drm]
 [<ffffffff8141821d>] ? sub_preempt_count+0x9d/0xd0
 [<ffffffffa0386f91>] i915_pci_probe+0x15/0x17 [i915]
 [<ffffffff8124481f>] local_pci_probe+0x5f/0xd0
 [<ffffffff81244f89>] pci_device_probe+0x119/0x120
 [<ffffffff812eccaa>] ? driver_sysfs_add+0x7a/0xb0
 [<ffffffff812ed003>] driver_probe_device+0xa3/0x290
 [<ffffffff812ed1f0>] ? __driver_attach+0x0/0xb0
 [<ffffffff812ed29b>] __driver_attach+0xab/0xb0
 [<ffffffff812ed1f0>] ? __driver_attach+0x0/0xb0
 [<ffffffff812ebd3e>] bus_for_each_dev+0x5e/0x90
 [<ffffffff812ecc2e>] driver_attach+0x1e/0x20
 [<ffffffff812ec6f2>] bus_add_driver+0xe2/0x320
 [<ffffffffa03aa000>] ? i915_init+0x0/0x96 [i915]
 [<ffffffff812ed536>] driver_register+0x76/0x140
 [<ffffffffa03aa000>] ? i915_init+0x0/0x96 [i915]
 [<ffffffff81245216>] __pci_register_driver+0x56/0xd0
 [<ffffffffa02f1264>] drm_pci_init+0xe4/0xf0 [drm]
 [<ffffffffa03aa000>] ? i915_init+0x0/0x96 [i915]
 [<ffffffffa02e84a8>] drm_init+0x58/0x70 [drm]
 [<ffffffffa03aa094>] i915_init+0x94/0x96 [i915]
 [<ffffffff81002194>] do_one_initcall+0x44/0x190
 [<ffffffff810a066b>] sys_init_module+0xcb/0x210
 [<ffffffff8100c012>] system_call_fastpath+0x16/0x1b
...

fb-test2 which reproduces above is available on kernel.org bug #26232.
To solve this issue, avoid calling lock_fb_info inside fb_set_suspend,
and move it out to where needed (callers of fb_set_suspend must call
lock_fb_info before if needed). So far, the only place which needs to
call lock_fb_info is store_fbstate, all other places which calls
fb_set_suspend are suspend/resume hooks that should not need the lock as
they should be run only when processes are already frozen in
suspend/resume.

References: https://bugzilla.kernel.org/show_bug.cgi?id&232
Signed-off-by: Herton Ronaldo Krzesinski <redacted>
---
 drivers/video/fbmem.c   |    4 +---
 drivers/video/fbsysfs.c |    3 +++
 2 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/drivers/video/fbmem.c b/drivers/video/fbmem.c
index 05d09ac..6747e06 100644
--- a/drivers/video/fbmem.c
+++ b/drivers/video/fbmem.c
@@ -1741,8 +1741,7 @@ void fb_set_suspend(struct fb_info *info, int state)
 {
 	struct fb_event event;
 
-	if (!lock_fb_info(info))
-		return;
+	WARN_ON(info->state = FBINFO_STATE_EXITING);
 	event.info = info;
 	if (state) {
 		fb_notifier_call_chain(FB_EVENT_SUSPEND, &event);
@@ -1751,7 +1750,6 @@ void fb_set_suspend(struct fb_info *info, int state)
 		info->state = FBINFO_STATE_RUNNING;
 		fb_notifier_call_chain(FB_EVENT_RESUME, &event);
 	}
-	unlock_fb_info(info);
 }
 
 /**
diff --git a/drivers/video/fbsysfs.c b/drivers/video/fbsysfs.c
index be361b5..2ba5370 100644
--- a/drivers/video/fbsysfs.c
+++ b/drivers/video/fbsysfs.c
@@ -401,9 +401,12 @@ static ssize_t store_fbstate(struct device *device,
 	if (state && state > 1)
 		return -EINVAL;
 
+	if (!lock_fb_info(fb_info))
+		return -ENODEV;
 	acquire_console_sem();
 	fb_set_suspend(fb_info, (int)state);
 	release_console_sem();
+	unlock_fb_info(fb_info);
 
 	return count;
 }
-- 
1.7.3.4
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help