[linux-next] mm/i915: i915_gemfs_init() NULL dereference

[linux-next] mm/i915: i915_gemfs_init() NULL dereference

[Sergey Senozhatsky]

Hi,

My laptop oopses early on with nothing on the screen;
after some debugging I managed to obtain a backtrace:

 BUG: kernel NULL pointer dereference, address: 0000000000000000
 #PF: supervisor instruction fetch in kernel mode
 #PF: error_code(0x0010) - not-present page
 PGD 0 P4D 0 
 Oops: 0010 [#1] PREEMPT SMP PTI
 RIP: 0010:0x0
 Code: Bad RIP value.
 [..]
 Call Trace:
  i915_gemfs_init+0x6e/0xa0 [i915]
  i915_gem_init_early+0x76/0x90 [i915]
  i915_driver_probe+0x30a/0x1640 [i915]
  ? kernfs_activate+0x5a/0x80
  ? kernfs_add_one+0xdd/0x130
  pci_device_probe+0x9e/0x110
  really_probe+0xce/0x230
  driver_probe_device+0x4b/0xc0
  device_driver_attach+0x4e/0x60
  __driver_attach+0x47/0xb0
  ? device_driver_attach+0x60/0x60
  bus_for_each_dev+0x61/0x90
  bus_add_driver+0x167/0x1b0
  driver_register+0x67/0xaa
  ? 0xffffffffc0522000
  do_one_initcall+0x37/0x13f
  ? kmem_cache_alloc+0x11a/0x150
  do_init_module+0x51/0x200
  __se_sys_init_module+0xef/0x100
  do_syscall_64+0x49/0x250
  entry_SYSCALL_64_after_hwframe+0x44/0xa9

RIP is at 0x00, which is never good

It sort of boils down to commit 144df3b288c4 (vfs: Convert ramfs,
shmem, tmpfs, devtmpfs, rootfs to use the new mount API), which
removed ->remount_fs from tmpfs' ops:

===
@@ -3736,7 +3849,6 @@ static const struct super_operations shmem_ops = {
        .destroy_inode  = shmem_destroy_inode,
 #ifdef CONFIG_TMPFS
        .statfs         = shmem_statfs,
-       .remount_fs     = shmem_remount_fs,
        .show_options   = shmem_show_options,
 #endif
        .evict_inode    = shmem_evict_inode,
===

So i915 init executes NULL

	get_fs_type("tmpfs");
	sb->s_op->remount_fs(sb, &flags, options);

For the time being the following (obvious and wrong) patch
at least boots -next:

---

diff --git a/drivers/gpu/drm/i915/gem/i915_gemfs.c b/drivers/gpu/drm/i915/gem/i915_gemfs.c
index 099f3397aada..1f95d9ea319a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gemfs.c
+++ b/drivers/gpu/drm/i915/gem/i915_gemfs.c
@@ -39,6 +39,9 @@ int i915_gemfs_init(struct drm_i915_private *i915)
 		int flags = 0;
 		int err;
 
+		if (!sb->s_op->remount_fs)
+			return -ENODEV;
+
 		err = sb->s_op->remount_fs(sb, &flags, options);
 		if (err) {
 			kern_unmount(gemfs);
---

	-ss

Re: [linux-next] mm/i915: i915_gemfs_init() NULL dereference

[Sergey Senozhatsky]

On (07/21/19 23:29), Sergey Senozhatsky wrote:
> 
>  BUG: kernel NULL pointer dereference, address: 0000000000000000
>  #PF: supervisor instruction fetch in kernel mode
>  #PF: error_code(0x0010) - not-present page
>  PGD 0 P4D 0 
>  Oops: 0010 [#1] PREEMPT SMP PTI
>  RIP: 0010:0x0
>  Code: Bad RIP value.
>  [..]
>  Call Trace:
>   i915_gemfs_init+0x6e/0xa0 [i915]
>   i915_gem_init_early+0x76/0x90 [i915]
>   i915_driver_probe+0x30a/0x1640 [i915]
>   ? kernfs_activate+0x5a/0x80
>   ? kernfs_add_one+0xdd/0x130
>   pci_device_probe+0x9e/0x110
>   really_probe+0xce/0x230
>   driver_probe_device+0x4b/0xc0
>   device_driver_attach+0x4e/0x60
>   __driver_attach+0x47/0xb0
>   ? device_driver_attach+0x60/0x60
>   bus_for_each_dev+0x61/0x90
>   bus_add_driver+0x167/0x1b0
>   driver_register+0x67/0xaa
>   ? 0xffffffffc0522000
>   do_one_initcall+0x37/0x13f
>   ? kmem_cache_alloc+0x11a/0x150
>   do_init_module+0x51/0x200
>   __se_sys_init_module+0xef/0x100
>   do_syscall_64+0x49/0x250
>   entry_SYSCALL_64_after_hwframe+0x44/0xa9


So "the new mount API" conversion probably looks like something below.
But I'm not 100% sure.

---
 drivers/gpu/drm/i915/gem/i915_gemfs.c | 32 +++++++++++++++++++++------
 1 file changed, 25 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gemfs.c b/drivers/gpu/drm/i915/gem/i915_gemfs.c
index 099f3397aada..2e365b26f8ee 100644
--- a/drivers/gpu/drm/i915/gem/i915_gemfs.c
+++ b/drivers/gpu/drm/i915/gem/i915_gemfs.c
@@ -7,12 +7,14 @@
 #include <linux/fs.h>
 #include <linux/mount.h>
 #include <linux/pagemap.h>
+#include <linux/fs_context.h>
 
 #include "i915_drv.h"
 #include "i915_gemfs.h"
 
 int i915_gemfs_init(struct drm_i915_private *i915)
 {
+	struct fs_context *fc;
 	struct file_system_type *type;
 	struct vfsmount *gemfs;
 
@@ -36,19 +38,35 @@ int i915_gemfs_init(struct drm_i915_private *i915)
 		struct super_block *sb = gemfs->mnt_sb;
 		/* FIXME: Disabled until we get W/A for read BW issue. */
 		char options[] = "huge=never";
-		int flags = 0;
 		int err;
 
-		err = sb->s_op->remount_fs(sb, &flags, options);
-		if (err) {
-			kern_unmount(gemfs);
-			return err;
-		}
+		fc = fs_context_for_reconfigure(sb->s_root, 0, 0);
+		if (IS_ERR(fc))
+			goto err;
+
+		if (!fc->ops->parse_monolithic)
+			goto err;
+
+		err = fc->ops->parse_monolithic(fc, options);
+		if (err)
+			goto err;
+
+		if (!fc->ops->reconfigure)
+			goto err;
+
+		err = fc->ops->reconfigure(fc);
+		if (err)
+			goto err;
 	}
 
 	i915->mm.gemfs = gemfs;
-
 	return 0;
+
+err:
+	pr_err("i915 gemfs init() failed\n");
+	put_fs_context(fc);
+	kern_unmount(gemfs);
+	return -EINVAL;
 }
 
 void i915_gemfs_fini(struct drm_i915_private *i915)
-- 
2.22.0

Re: [Intel-gfx] [linux-next] mm/i915: i915_gemfs_init() NULL dereference

[Chris Wilson]

Quoting Sergey Senozhatsky (2019-07-31 17:48:29)
> @@ -36,19 +38,35 @@ int i915_gemfs_init(struct drm_i915_private *i915)
>                 struct super_block *sb = gemfs->mnt_sb;
>                 /* FIXME: Disabled until we get W/A for read BW issue. */
>                 char options[] = "huge=never";
> -               int flags = 0;
>                 int err;
>  
> -               err = sb->s_op->remount_fs(sb, &flags, options);
> -               if (err) {
> -                       kern_unmount(gemfs);
> -                       return err;
> -               }
> +               fc = fs_context_for_reconfigure(sb->s_root, 0, 0);
> +               if (IS_ERR(fc))
> +                       goto err;
> +
> +               if (!fc->ops->parse_monolithic)
> +                       goto err;
> +
> +               err = fc->ops->parse_monolithic(fc, options);
> +               if (err)
> +                       goto err;
> +
> +               if (!fc->ops->reconfigure)

It would be odd for fs_context_for_reconfigure() to allow creation of a
context if that context couldn't perform a reconfigre, nevertheless that
seems to be the case.

> +                       goto err;
> +
> +               err = fc->ops->reconfigure(fc);
> +               if (err)
> +                       goto err;

Only thing that stands out is that we should put_fs_context() here as
well. I guess it's better than poking at the SB_INFO directly ourselves.

I think though we shouldn't bail if we can't change the thp setting, and
just accept whatever with a warning.

Looks like the API is already available in dinq, so we can apply this
ahead of the next merge window.
-Chris

Re: [Intel-gfx] [linux-next] mm/i915: i915_gemfs_init() NULL dereference

[Sergey Senozhatsky]

On (08/01/19 18:30), Chris Wilson wrote:
> Quoting Sergey Senozhatsky (2019-07-31 17:48:29)
> > @@ -36,19 +38,35 @@ int i915_gemfs_init(struct drm_i915_private *i915)
[..]
> > +               if (!fc->ops->parse_monolithic)
> > +                       goto err;
> > +
> > +               err = fc->ops->parse_monolithic(fc, options);
> > +               if (err)
> > +                       goto err;
> > +
> > +               if (!fc->ops->reconfigure)
> 
> It would be odd for fs_context_for_reconfigure() to allow creation of a
> context if that context couldn't perform a reconfigre, nevertheless that
> seems to be the case.

Well, I kept those checks just because fs/ code does the same.

E.g. fs/super.c

	reconfigure_super()
	{
		if (fc->ops->reconfigure)
			fc->ops->reconfigure(fc)
	}

	do_emergency_remount_callback()
	{
		fc = fs_context_for_reconfigure();
		reconfigure_super(fc);
	}

> > +                       goto err;
> > +
> > +               err = fc->ops->reconfigure(fc);
> > +               if (err)
> > +                       goto err;
> 
> Only thing that stands out is that we should put_fs_context() here as
> well.

Oh... Indeed, somehow I forgot to put_fs_context().

> I guess it's better than poking at the SB_INFO directly ourselves.
> I think though we shouldn't bail if we can't change the thp setting, and
> just accept whatever with a warning.

OK.

> Looks like the API is already available in dinq, so we can apply this
> ahead of the next merge window.

OK, will cook a formal patch then. Thanks!

	-ss