From patchwork Tue May 10 12:47:42 2011 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Tim Gardner X-Patchwork-Id: 774222 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by demeter1.kernel.org (8.14.4/8.14.3) with ESMTP id p4AClT09014614 for ; Tue, 10 May 2011 12:47:59 GMT Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752884Ab1EJMr7 (ORCPT ); Tue, 10 May 2011 08:47:59 -0400 Received: from mail.tpi.com ([70.99.223.143]:3469 "EHLO mail.tpi.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752496Ab1EJMr6 (ORCPT ); Tue, 10 May 2011 08:47:58 -0400 Received: from [127.0.0.1] (mail.tpi.com [70.99.223.143]) by mail.tpi.com (Postfix) with ESMTP id C14E12C09BC; Tue, 10 May 2011 05:47:43 -0700 (PDT) Message-ID: <4DC933EE.70409@canonical.com> Date: Tue, 10 May 2011 14:47:42 +0200 From: Tim Gardner User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.15) Gecko/20110419 Thunderbird/3.1.9 MIME-Version: 1.0 To: Jack Stone CC: linux-fbdev@vger.kernel.org, lethal@linux-sh.org, linux-kernel@vger.kernel.org, Andy Whitcroft , Leann Ogasawara Subject: [PATCH V2] fbcon -- fix race between open and removal of framebuffers References: <1304617307-7389-1-git-send-email-tim.gardner@canonical.com> <1304617307-7389-2-git-send-email-tim.gardner@canonical.com> <4DC30FFA.9030708@fastmail.fm> In-Reply-To: <4DC30FFA.9030708@fastmail.fm> Sender: linux-fbdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-fbdev@vger.kernel.org X-Greylist: IP, sender and recipient auto-whitelisted, not delayed by milter-greylist-4.2.6 (demeter1.kernel.org [140.211.167.41]); Tue, 10 May 2011 12:47:59 +0000 (UTC) On 05/05/2011 11:00 PM, Jack Stone wrote: > On 05/05/2011 18:41, tim.gardner@canonical.com wrote: >> +static struct fb_info *get_framebuffer_info(int idx) >> +__acquires(®istered_lock) >> +__releases(®istered_lock) >> +{ >> + struct fb_info *fb_info; >> + >> + spin_lock(®istered_lock); >> + fb_info = registered_fb[idx]; >> + fb_info->ref_count++; >> + spin_unlock(®istered_lock); >> + >> + return fb_info; >> +} >> + >> static int >> fb_open(struct inode *inode, struct file *file) >> __acquires(&info->lock) >> @@ -1363,13 +1421,17 @@ __releases(&info->lock) >> >> if (fbidx>= FB_MAX) >> return -ENODEV; >> - info = registered_fb[fbidx]; >> + info = get_framebuffer_info(fbidx); >> if (!info) >> request_module("fb%d", fbidx); >> - info = registered_fb[fbidx]; >> + info = get_framebuffer_info(fbidx); >> if (!info) >> return -ENODEV; > > If the first get_framebuffer_info succeeds don't you up the ref count > twice? Shouldn't this be: > > info = get_framebuffer_info(fbidx); > if (!info) { > request_module("fb%d", fbidx); > info = get_framebuffer_info(fbidx); > } > if (!info) > return -ENODEV; > > Thanks, > > Jack Good catch. See attached. rtg From cefd70cfbb798cca488d6e97b25f41158f222afc Mon Sep 17 00:00:00 2001 From: Andy Whitcroft Date: Thu, 29 Jul 2010 16:48:20 +0100 Subject: [PATCH] fbcon -- fix race between open and removal of framebuffers Currently there is no locking for updates to the registered_fb list. This allows an open through /dev/fbN to pick up a registered framebuffer pointer in parallel with it being released, as happens when a conflicting framebuffer is ejected or on module unload. There is also no reference counting on the framebuffer descriptor which is referenced from all open files, leading to references to released or reused memory to persist on these open files. This patch adds a reference count to the framebuffer descriptor to prevent it from being released until after all pending opens are closed. This allows the pending opens to detect the closed status and unmap themselves. It also adds locking to the framebuffer lookup path, locking it against the removal path such that it is possible to atomically lookup and take a reference to the descriptor. It also adds locking to the read and write paths which currently could access the framebuffer descriptor after it has been freed. Finally it moves the device to FBINFO_STATE_REMOVED to indicate that all access should be errored for this device. Signed-off-by: Andy Whitcroft Acked-by: Stefan Bader Signed-off-by: Leann Ogasawara Signed-off-by: Tim Gardner --- drivers/video/fbmem.c | 135 ++++++++++++++++++++++++++++++++++++++----------- include/linux/fb.h | 2 + 2 files changed, 107 insertions(+), 30 deletions(-) diff --git a/drivers/video/fbmem.c b/drivers/video/fbmem.c index e0c2284..ea333dd 100644 --- a/drivers/video/fbmem.c +++ b/drivers/video/fbmem.c @@ -42,6 +42,8 @@ #define FBPIXMAPSIZE (1024 * 8) +/* Protects the registered framebuffer list and count. */ +static DEFINE_SPINLOCK(registered_lock); struct fb_info *registered_fb[FB_MAX] __read_mostly; int num_registered_fb __read_mostly; @@ -694,9 +696,7 @@ 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; @@ -705,19 +705,28 @@ fb_read(struct file *file, char __user *buf, size_t count, loff_t *ppos) if (!info || ! info->screen_base) return -ENODEV; - if (info->state != FBINFO_STATE_RUNNING) - return -EPERM; + if (!lock_fb_info(info)) + return -ENODEV; + + if (info->state != FBINFO_STATE_RUNNING) { + err = -EPERM; + goto out_fb_info; + } - if (info->fbops->fb_read) - return info->fbops->fb_read(info, buf, count, ppos); + if (info->fbops->fb_read) { + err = info->fbops->fb_read(info, buf, count, ppos); + goto out_fb_info; + } 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 out_fb_info; + } if (count >= total_size) count = total_size; @@ -727,8 +736,10 @@ 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 out_fb_info; + } src = (u8 __iomem *) (info->screen_base + p); @@ -751,19 +762,21 @@ fb_read(struct file *file, char __user *buf, size_t count, loff_t *ppos) cnt += c; count -= c; } + if (!err) + err = cnt; kfree(buffer); +out_fb_info: + unlock_fb_info(info); - return (err) ? err : cnt; + 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; @@ -772,8 +785,13 @@ fb_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos) if (!info || !info->screen_base) return -ENODEV; - if (info->state != FBINFO_STATE_RUNNING) - return -EPERM; + if (!lock_fb_info(info)) + return -ENODEV; + + if (info->state != FBINFO_STATE_RUNNING) { + err = -EPERM; + goto out_fb_info; + } if (info->fbops->fb_write) return info->fbops->fb_write(info, buf, count, ppos); @@ -783,8 +801,10 @@ fb_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos) if (total_size == 0) total_size = info->fix.smem_len; - if (p > total_size) - return -EFBIG; + if (p > total_size) { + err = -EFBIG; + goto out_fb_info; + } if (count > total_size) { err = -EFBIG; @@ -800,8 +820,10 @@ 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 out_fb_info; + } dst = (u8 __iomem *) (info->screen_base + p); @@ -825,10 +847,14 @@ fb_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos) cnt += c; count -= c; } + if (cnt) + err = cnt; kfree(buffer); +out_fb_info: + unlock_fb_info(info); - return (cnt) ? cnt : err; + return err; } int @@ -1303,8 +1329,7 @@ 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 * const info = file->private_data; struct fb_ops *fb = info->fbops; unsigned long off; unsigned long start; @@ -1316,6 +1341,11 @@ fb_mmap(struct file *file, struct vm_area_struct * vma) if (!fb) return -ENODEV; mutex_lock(&info->mm_lock); + if (info->state == FBINFO_STATE_REMOVED) { + mutex_unlock(&info->mm_lock); + return -ENODEV; + } + if (fb->fb_mmap) { int res; res = fb->fb_mmap(info, vma); @@ -1352,6 +1382,34 @@ fb_mmap(struct file *file, struct vm_area_struct * vma) return 0; } +static struct fb_info *get_framebuffer_info(int idx) +__acquires(®istered_lock) +__releases(®istered_lock) +{ + struct fb_info *fb_info; + + spin_lock(®istered_lock); + fb_info = registered_fb[idx]; + fb_info->ref_count++; + spin_unlock(®istered_lock); + + return fb_info; +} + +static void put_framebuffer_info(struct fb_info *fb_info) +__acquires(®istered_lock) +__releases(®istered_lock) +{ + int keep; + + spin_lock(®istered_lock); + keep = --fb_info->ref_count; + spin_unlock(®istered_lock); + + if (!keep && fb_info->fbops->fb_destroy) + fb_info->fbops->fb_destroy(fb_info); +} + static int fb_open(struct inode *inode, struct file *file) __acquires(&info->lock) @@ -1363,13 +1421,18 @@ __releases(&info->lock) if (fbidx >= FB_MAX) return -ENODEV; - info = registered_fb[fbidx]; - if (!info) + info = get_framebuffer_info(fbidx); + if (!info) { request_module("fb%d", fbidx); - info = registered_fb[fbidx]; + info = get_framebuffer_info(fbidx); + } if (!info) return -ENODEV; mutex_lock(&info->lock); + if (info->state == FBINFO_STATE_REMOVED) { + res = -ENODEV; + goto out; + } if (!try_module_get(info->fbops->owner)) { res = -ENODEV; goto out; @@ -1386,6 +1449,8 @@ __releases(&info->lock) #endif out: mutex_unlock(&info->lock); + if (res) + put_framebuffer_info(info); return res; } @@ -1401,6 +1466,7 @@ __releases(&info->lock) info->fbops->fb_release(info,1); module_put(info->fbops->owner); mutex_unlock(&info->lock); + put_framebuffer_info(info); return 0; } @@ -1549,6 +1615,7 @@ register_framebuffer(struct fb_info *fb_info) fb_info->node = i; mutex_init(&fb_info->lock); mutex_init(&fb_info->mm_lock); + fb_info->ref_count = 1; fb_info->dev = device_create(fb_class, fb_info->device, MKDEV(FB_MAJOR, i), NULL, "fb%d", i); @@ -1592,7 +1659,6 @@ register_framebuffer(struct fb_info *fb_info) return 0; } - /** * unregister_framebuffer - releases a frame buffer device * @fb_info: frame buffer info structure @@ -1627,6 +1693,16 @@ unregister_framebuffer(struct fb_info *fb_info) return -ENODEV; event.info = fb_info; ret = fb_notifier_call_chain(FB_EVENT_FB_UNBIND, &event); + if (!ret) { + mutex_lock(&fb_info->mm_lock); + /* + * We must prevent any operations for this transition, we + * already have info->lock so grab the info->mm_lock to hold + * the remainder. + */ + fb_info->state = FBINFO_STATE_REMOVED; + mutex_unlock(&fb_info->mm_lock); + } unlock_fb_info(fb_info); if (ret) { @@ -1646,8 +1722,7 @@ unregister_framebuffer(struct fb_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); + put_framebuffer_info(fb_info); done: return ret; } diff --git a/include/linux/fb.h b/include/linux/fb.h index df728c1..60de3fa 100644 --- a/include/linux/fb.h +++ b/include/linux/fb.h @@ -834,6 +834,7 @@ struct fb_tile_ops { struct fb_info { int node; int flags; + int ref_count; struct mutex lock; /* Lock for open/release/ioctl funcs */ struct mutex mm_lock; /* Lock for fb_mmap and smem_* fields */ struct fb_var_screeninfo var; /* Current var */ @@ -873,6 +874,7 @@ struct fb_info { void *pseudo_palette; /* Fake palette of 16 colors */ #define FBINFO_STATE_RUNNING 0 #define FBINFO_STATE_SUSPENDED 1 +#define FBINFO_STATE_REMOVED 2 u32 state; /* Hardware state i.e suspend */ void *fbcon_par; /* fbcon use-only private area */ /* From here on everything is device dependent */ -- 1.7.4.1