2020-03-28 02:00:03

by Chen Wandun

[permalink] [raw]
Subject: [PATCH next] vt: fix a warning when kmalloc alloc large memory

If the memory size that use kmalloc() to allocate exceed MAX_ORDER pages,
it will hit the WARN_ON_ONCE(!(gfp_mask & __GFP_NOWARN)), so add memory
allocation flag __GFP_NOWARN to silence a warning, othervise, it will
cause panic if panic_on_warn is enable.

The calltrace is:

WARNING: CPU: 1 PID: 6298 at mm/page_alloc.c:4713 __alloc_pages_nodemask+0x339/0x7d0 mm/page_alloc.c:4713

Call Trace:
__alloc_pages_nodemask+0x339/0x7d0 mm/page_alloc.c:4713
alloc_pages_current+0xac/0x1e0 mm/mempolicy.c:2211
alloc_pages include/linux/gfp.h:532 [inline]
kmalloc_order+0x21/0xf0 mm/slab_common.c:1324
kmalloc_order_trace+0x18/0x150 mm/slab_common.c:1340
kmalloc include/linux/slab.h:560 [inline]
vc_uniscr_alloc+0x2b/0xb0 drivers/tty/vt/vt.c:353
vc_do_resize+0x319/0x12b0 drivers/tty/vt/vt.c:1203
vt_ioctl+0xa5f/0x29a0 drivers/tty/vt/vt_ioctl.c:840
tty_ioctl+0x27d/0x1420 drivers/tty/tty_io.c:2660
vfs_ioctl fs/ioctl.c:47 [inline]
ksys_ioctl+0xee/0x120 fs/ioctl.c:763
__do_sys_ioctl fs/ioctl.c:772 [inline]
__se_sys_ioctl fs/ioctl.c:770 [inline]
__x64_sys_ioctl+0x6f/0xb0 fs/ioctl.c:770
do_syscall_64+0xa1/0x540 arch/x86/entry/common.c:294
entry_SYSCALL_64_after_hwframe+0x49/0xbe

Fixes: d8ae72427187 ("vt: preserve unicode values corresponding to screen characters")
Signed-off-by: Chen Wandun <[email protected]>
---
drivers/tty/vt/vt.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/tty/vt/vt.c b/drivers/tty/vt/vt.c
index bbc26d73209a..2b000b31a351 100644
--- a/drivers/tty/vt/vt.c
+++ b/drivers/tty/vt/vt.c
@@ -350,7 +350,7 @@ static struct uni_screen *vc_uniscr_alloc(unsigned int cols, unsigned int rows)
/* allocate everything in one go */
memsize = cols * rows * sizeof(char32_t);
memsize += rows * sizeof(char32_t *);
- p = kmalloc(memsize, GFP_KERNEL);
+ p = kmalloc(memsize, GFP_KERNEL | __GFP_NOWARN);
if (!p)
return NULL;

--
2.17.1


2020-03-28 02:55:56

by Nicolas Pitre

[permalink] [raw]
Subject: Re: [PATCH next] vt: fix a warning when kmalloc alloc large memory

On Sat, 28 Mar 2020, Chen Wandun wrote:

> If the memory size that use kmalloc() to allocate exceed MAX_ORDER pages,
> it will hit the WARN_ON_ONCE(!(gfp_mask & __GFP_NOWARN)), so add memory
> allocation flag __GFP_NOWARN to silence a warning, othervise, it will
> cause panic if panic_on_warn is enable.

Wow! How do you manage that? You need something like a 1024x1024 text
screen to get such a big memory allocation.

Still, GFP_NOWARN is not a proper fix. This kmalloc() should rather be
replaced by vmalloc(), and corresponding kfree() should then be vfree().
There is no need for this allocation to be physically contiguous.


Nicolas

2020-03-28 03:13:54

by Adam Borowski

[permalink] [raw]
Subject: Re: [PATCH next] vt: fix a warning when kmalloc alloc large memory

On Fri, Mar 27, 2020 at 10:55:14PM -0400, Nicolas Pitre wrote:
> On Sat, 28 Mar 2020, Chen Wandun wrote:
>
> > If the memory size that use kmalloc() to allocate exceed MAX_ORDER pages,
> > it will hit the WARN_ON_ONCE(!(gfp_mask & __GFP_NOWARN)), so add memory
> > allocation flag __GFP_NOWARN to silence a warning, othervise, it will
> > cause panic if panic_on_warn is enable.
>
> Wow! How do you manage that? You need something like a 1024x1024 text
> screen to get such a big memory allocation.

ioctl(VT_RESIZE) allows up to 32767x32767, unprivileged for a local user.
That's 4GB per console.


Meow!
--
⢀⣴⠾⠻⢶⣦⠀
⣾⠁⢠⠒⠀⣿⡁ in the beginning was the boot and root floppies and they were good.
⢿⡄⠘⠷⠚⠋⠀ -- <willmore> on #linux-sunxi
⠈⠳⣄⠀⠀⠀⠀

2020-03-28 16:21:43

by Nicolas Pitre

[permalink] [raw]
Subject: Re: [PATCH next] vt: fix a warning when kmalloc alloc large memory

On Sat, 28 Mar 2020, Adam Borowski wrote:

> On Fri, Mar 27, 2020 at 10:55:14PM -0400, Nicolas Pitre wrote:
> > On Sat, 28 Mar 2020, Chen Wandun wrote:
> >
> > > If the memory size that use kmalloc() to allocate exceed MAX_ORDER pages,
> > > it will hit the WARN_ON_ONCE(!(gfp_mask & __GFP_NOWARN)), so add memory
> > > allocation flag __GFP_NOWARN to silence a warning, othervise, it will
> > > cause panic if panic_on_warn is enable.
> >
> > Wow! How do you manage that? You need something like a 1024x1024 text
> > screen to get such a big memory allocation.
>
> ioctl(VT_RESIZE) allows up to 32767x32767, unprivileged for a local user.
> That's 4GB per console.

Right. And that's ridiculous. This ought to be limited to something much
saner.


Nicolas

2020-03-28 20:52:50

by Nicolas Pitre

[permalink] [raw]
Subject: Re: [PATCH next] vt: fix a warning when kmalloc alloc large memory

On Sat, 28 Mar 2020, Adam Borowski wrote:

> On Fri, Mar 27, 2020 at 10:55:14PM -0400, Nicolas Pitre wrote:
> > On Sat, 28 Mar 2020, Chen Wandun wrote:
> >
> > > If the memory size that use kmalloc() to allocate exceed MAX_ORDER pages,
> > > it will hit the WARN_ON_ONCE(!(gfp_mask & __GFP_NOWARN)), so add memory
> > > allocation flag __GFP_NOWARN to silence a warning, othervise, it will
> > > cause panic if panic_on_warn is enable.
> >
> > Wow! How do you manage that? You need something like a 1024x1024 text
> > screen to get such a big memory allocation.
>
> ioctl(VT_RESIZE) allows up to 32767x32767, unprivileged for a local user.
> That's 4GB per console.

In fact that's not exactly true. The code has this protection:

if (new_screen_size > (4 << 20))
return -EINVAL;

The problem is with the unicode screen content whose buffer is larger
than the legacy glyph buffer. Still, the above test is a bit iffy as it
depends on the default MAX_ORDER value which is configurable.


Nicolas