lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:	Wed, 18 Dec 2013 15:21:24 +0100
From:	Takashi Iwai <tiwai@...e.de>
To:	David Herrmann <dh.herrmann@...il.com>
Cc:	linux-kernel@...r.kernel.org,
	Stephen Warren <swarren@...dotorg.org>,
	the arch/x86 maintainers <x86@...nel.org>,
	linux-fbdev@...r.kernel.org,
	Geert Uytterhoeven <geert@...ux-m68k.org>,
	Ingo Molnar <mingo@...nel.org>
Subject: Re: [PATCH v2] x86: sysfb: remove sysfb when probing real hw

At Wed, 18 Dec 2013 14:50:11 +0100,
David Herrmann wrote:
> 
> With CONFIG_X86_SYSFB=y, probing real hw-drivers may result in
> resource-conflicts and drivers will refuse to load. A call to
> request_mem_region() will fail, if the region overlaps with the mem-region
> used by simplefb. The common desktop DRM drivers (intel, nouveau, radeon)
> are not affected as they don't reserve their resources, but some others
> do, including (nvidiafb, cirrus, ..).
> 
> The problem is that we add an IORESOURCE_MEM to the simple-framebuffer
> platform-device during bootup but never release it. Probing simplefb on
> this platform-device is fine, but the handover to real-hw via
> remove_conflicting_framebuffers() will only unload the fbdev driver, but
> keep the platform-device around. Thus, if the real hw-driver calls
> request_mem_region() and friends on the same PCI region, we will get a
> resource conflict and most drivers refuse to load. Users will see
> errors like:
>   "nvidiafb: cannot reserve video memory at <xyz>"
> 
> vesafb and efifb don't store any resources, so disabling CONFIG_X86_SYSFB
> and falling back to those drivers will avoid the bug. With sysfb enabled,
> we need to properly unload the simple-framebuffer devices before probing
> real hw-drivers.
> 
> This patch adds sysfb_unregister() for that purpose. It can be called from
> any context (except from the platform-device ->probe and ->remove callback
> path), synchronously unloads any global sysfb and prevents new sysfbs from
> getting registered. Thus, you can call it even before any sysfb has been
> loaded. Note that for now we only do that for simple-framebuffer devices,
> as efi/vesa-framebuffer platform-drivers lack ->remove() callbacks.
> They're not affected by the resource-conflicts, so we can fix those later.
> 
> This also changes remove_conflicting_framebuffer() to call this helper
> *before* trying its heuristic to remove conflicting drivers. This way, we
> unload sysfbs properly on any conflict. But to avoid dead-locks in
> register_framebuffer(), we must not call sysfb_unregister() for
> framebuffers probing on sysfb devices. Hence, we simply remove any
> aperture from simplefb and we're good to go. simplefb is unregistered by
> sysfb_unregister() now, so no reason to keep the apertures (on non-x86,
> there currently is no handover from simplefb, so we're fine. If it's
> added, they need to provide something like sysfb_unregister() too)
> 
> Signed-off-by: David Herrmann <dh.herrmann@...il.com>
> ---
> v2:
>  - remove simplefb apertures to avoid dead-lock
>  - skip sysfb_unregister() if no apertures are set
>  - merge sysfb_alloc() into sysfb_register()
>  - simplify sysfb_unregister() logic
>  - call sysfb_unregister() even if SYSFB=n but X86=y
> 
> Tested with efifb and simplefb + i915 handover on x86.

The patch worked fine on QEMU with cirrus driver.

I'd like to see comments, as suggested in my previous mail, in a
couple of places.  Other than that, the patch looks good to me.

So, feel free to my tested-by and reviewed-by tags:

	Reviewed-by: Takashi Iwai <tiwai@...e.de>
	Tested-by: Takashi Iwai <tiwai@...e.de>


thanks,

Takashi

> 
> Thanks
> David
> 
>  arch/x86/include/asm/sysfb.h     |  6 ++++
>  arch/x86/kernel/sysfb.c          | 63 ++++++++++++++++++++++++++++++++++++++++
>  arch/x86/kernel/sysfb_simplefb.c |  9 ++----
>  drivers/video/fbmem.c            | 19 ++++++++++++
>  drivers/video/simplefb.c         |  8 -----
>  5 files changed, 90 insertions(+), 15 deletions(-)
> 
> diff --git a/arch/x86/include/asm/sysfb.h b/arch/x86/include/asm/sysfb.h
> index 2aeb3e2..10d719d 100644
> --- a/arch/x86/include/asm/sysfb.h
> +++ b/arch/x86/include/asm/sysfb.h
> @@ -11,6 +11,7 @@
>   * any later version.
>   */
>  
> +#include <linux/fb.h>
>  #include <linux/kernel.h>
>  #include <linux/platform_data/simplefb.h>
>  #include <linux/screen_info.h>
> @@ -59,6 +60,11 @@ struct efifb_dmi_info {
>  	int flags;
>  };
>  
> +__init int sysfb_register(const char *name, int id,
> +			  const struct resource *res, unsigned int res_num,
> +			  const void *data, size_t data_size);

In most cases, we declare like
	int __init sysfb_register(...

> +void sysfb_unregister(const struct apertures_struct *apert, bool primary);
> +
>  #ifdef CONFIG_EFI
>  
>  extern struct efifb_dmi_info efifb_dmi_list[];
> diff --git a/arch/x86/kernel/sysfb.c b/arch/x86/kernel/sysfb.c
> index 193ec2c..9d8da8d 100644
> --- a/arch/x86/kernel/sysfb.c
> +++ b/arch/x86/kernel/sysfb.c
> @@ -33,11 +33,74 @@
>  #include <linux/init.h>
>  #include <linux/kernel.h>
>  #include <linux/mm.h>
> +#include <linux/mutex.h>
>  #include <linux/platform_data/simplefb.h>
>  #include <linux/platform_device.h>
>  #include <linux/screen_info.h>
>  #include <asm/sysfb.h>
>  
> +static DEFINE_MUTEX(sysfb_lock);
> +static struct platform_device *sysfb_dev;
> +
> +__init int sysfb_register(const char *name, int id,
> +			  const struct resource *res, unsigned int res_num,
> +			  const void *data, size_t data_size)
> +{
> +	struct platform_device *pd;
> +	int ret = 0;
> +
> +	mutex_lock(&sysfb_lock);
> +	if (!sysfb_dev) {
> +		pd = platform_device_register_resndata(NULL, name, id,
> +						       res, res_num,
> +						       data, data_size);
> +		if (IS_ERR(pd))
> +			ret = PTR_ERR(pd);
> +		else
> +			sysfb_dev = pd;
> +	}
> +	mutex_unlock(&sysfb_lock);
> +
> +	return ret;
> +}
> +
> +static bool sysfb_match(const struct apertures_struct *apert, bool primary)
> +{
> +	struct screen_info *si = &screen_info;
> +	unsigned int i;
> +	const struct aperture *a;
> +
> +	if (!apert || primary)
> +		return true;
> +
> +	for (i = 0; i < apert->count; ++i) {
> +		a = &apert->ranges[i];
> +		if (a->base >= si->lfb_base &&
> +		    a->base < si->lfb_base + ((u64)si->lfb_size << 16))
> +			return true;
> +		if (si->lfb_base >= a->base &&
> +		    si->lfb_base < a->base + a->size)
> +			return true;
> +	}
> +
> +	return false;
> +}
> +
> +/* unregister the sysfb and prevent new sysfbs from getting registered */
> +void sysfb_unregister(const struct apertures_struct *apert, bool primary)
> +{
> +	mutex_lock(&sysfb_lock);
> +	if (!IS_ERR(sysfb_dev) && sysfb_dev) {
> +		if (sysfb_match(apert, primary)) {
> +			platform_device_unregister(sysfb_dev);
> +			sysfb_dev = ERR_PTR(-EALREADY);
> +		}
> +	} else {
> +		sysfb_dev = ERR_PTR(-EALREADY);
> +	}
> +	mutex_unlock(&sysfb_lock);
> +}
> +
>  static __init int sysfb_init(void)
>  {
>  	struct screen_info *si = &screen_info;
> diff --git a/arch/x86/kernel/sysfb_simplefb.c b/arch/x86/kernel/sysfb_simplefb.c
> index 86179d4..a760d47 100644
> --- a/arch/x86/kernel/sysfb_simplefb.c
> +++ b/arch/x86/kernel/sysfb_simplefb.c
> @@ -64,7 +64,6 @@ __init bool parse_mode(const struct screen_info *si,
>  __init int create_simplefb(const struct screen_info *si,
>  			   const struct simplefb_platform_data *mode)
>  {
> -	struct platform_device *pd;
>  	struct resource res;
>  	unsigned long len;
>  
> @@ -86,10 +85,6 @@ __init int create_simplefb(const struct screen_info *si,
>  	if (res.end <= res.start)
>  		return -EINVAL;
>  
> -	pd = platform_device_register_resndata(NULL, "simple-framebuffer", 0,
> -					       &res, 1, mode, sizeof(*mode));
> -	if (IS_ERR(pd))
> -		return PTR_ERR(pd);
> -
> -	return 0;
> +	return sysfb_register("simple-framebuffer", 0, &res, 1, mode,
> +			      sizeof(*mode));
>  }
> diff --git a/drivers/video/fbmem.c b/drivers/video/fbmem.c
> index 010d191..590a46a 100644
> --- a/drivers/video/fbmem.c
> +++ b/drivers/video/fbmem.c
> @@ -35,6 +35,9 @@
>  
>  #include <asm/fb.h>
>  
> +#if IS_ENABLED(CONFIG_X86)
> +#include <asm/sysfb.h>
> +#endif
>  
>      /*
>       *  Frame buffer device initialization and setup routines
> @@ -1604,6 +1607,17 @@ static void do_remove_conflicting_framebuffers(struct apertures_struct *a,
>  	}
>  }
>  
> +static void remove_conflicting_sysfb(const struct apertures_struct *apert,
> +				     bool primary)
> +{
> +	if (!apert)
> +		return;
> +
> +#if IS_ENABLED(CONFIG_X86)
> +	sysfb_unregister(apert, primary);
> +#endif
> +}
> +
>  static int do_register_framebuffer(struct fb_info *fb_info)
>  {
>  	int i;
> @@ -1742,6 +1756,8 @@ EXPORT_SYMBOL(unlink_framebuffer);
>  void remove_conflicting_framebuffers(struct apertures_struct *a,
>  				     const char *name, bool primary)
>  {
> +	remove_conflicting_sysfb(a, primary);
> +
>  	mutex_lock(&registration_lock);
>  	do_remove_conflicting_framebuffers(a, name, primary);
>  	mutex_unlock(&registration_lock);
> @@ -1762,6 +1778,9 @@ register_framebuffer(struct fb_info *fb_info)
>  {
>  	int ret;
>  
> +	remove_conflicting_sysfb(fb_info->apertures,
> +				 fb_is_primary_device(fb_info));
> +
>  	mutex_lock(&registration_lock);
>  	ret = do_register_framebuffer(fb_info);
>  	mutex_unlock(&registration_lock);
> diff --git a/drivers/video/simplefb.c b/drivers/video/simplefb.c
> index 210f3a0..9f4a0cf 100644
> --- a/drivers/video/simplefb.c
> +++ b/drivers/video/simplefb.c
> @@ -209,14 +209,6 @@ static int simplefb_probe(struct platform_device *pdev)
>  	info->var.blue = params.format->blue;
>  	info->var.transp = params.format->transp;
>  
> -	info->apertures = alloc_apertures(1);
> -	if (!info->apertures) {
> -		framebuffer_release(info);
> -		return -ENOMEM;
> -	}
> -	info->apertures->ranges[0].base = info->fix.smem_start;
> -	info->apertures->ranges[0].size = info->fix.smem_len;
> -
>  	info->fbops = &simplefb_ops;
>  	info->flags = FBINFO_DEFAULT | FBINFO_MISC_FIRMWARE;
>  	info->screen_base = ioremap_wc(info->fix.smem_start,
> -- 
> 1.8.5.1
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ