On 2020/7/30 19:04, Jiri Slaby wrote:

On 13. 07. 20, 12:57, Yang Yingliang wrote:

quoted

I got a slab-out-of-bounds report when I doing fuzz test.

[  334.989515] =================================
[  334.989577] BUG: KASAN: slab-out-of-bounds in vgacon_scroll+0x57a/0x8ed
[  334.989588] Write of size 1766 at addr ffff8883de69ff3e by task test/2658
[  334.989593]
[  334.989608] CPU: 3 PID: 2658 Comm: test Not tainted 5.7.0-rc5-00005-g152036d1379f #789
[  334.989617] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.1-0-ga5cab58e9a3f-prebuilt.qemu.org 04/01/2014
[  334.989624] Call Trace:
[  334.989646]  dump_stack+0xe4/0x14e
[  334.989676]  print_address_description.constprop.5+0x3f/0x60
[  334.989699]  ? vgacon_scroll+0x57a/0x8ed
[  334.989710]  __kasan_report.cold.8+0x92/0xaf
[  334.989735]  ? vgacon_scroll+0x57a/0x8ed
[  334.989761]  kasan_report+0x37/0x50
[  334.989789]  check_memory_region+0x1c1/0x1e0
[  334.989806]  memcpy+0x38/0x60
[  334.989824]  vgacon_scroll+0x57a/0x8ed
[  334.989876]  con_scroll+0x4ef/0x5e0

...

quoted

Because vgacon_scrollback_cur->tail plus memcpy size is greater than
vgacon_scrollback_cur->size. Fix this by checking the memcpy size.

Reported-by: Hulk Robot <redacted>
Signed-off-by: Yang Yingliang <redacted>
---
  drivers/video/console/vgacon.c | 11 ++++++++---
  1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/drivers/video/console/vgacon.c b/drivers/video/console/vgacon.c
index 998b0de1812f..b51ffb9a208d 100644
--- a/drivers/video/console/vgacon.c
+++ b/drivers/video/console/vgacon.c

@@ -243,6 +243,7 @@ static void vgacon_scrollback_startup(void)
  static void vgacon_scrollback_update(struct vc_data *c, int t, int count)
  {
  	void *p;
+	int size;
  
  	if (!vgacon_scrollback_cur->data || !vgacon_scrollback_cur->size ||
  	    c->vc_num != fg_console)

@@ -251,13 +252,17 @@ static void vgacon_scrollback_update(struct vc_data *c, int t, int count)
  	p = (void *) (c->vc_origin + t * c->vc_size_row);
  
  	while (count--) {
+		size = vgacon_scrollback_cur->size - vgacon_scrollback_cur->tail;
+		if (size > c->vc_size_row)
+			size = c->vc_size_row;
+
  		scr_memcpyw(vgacon_scrollback_cur->data +
  			    vgacon_scrollback_cur->tail,
-			    p, c->vc_size_row);
+			    p, size);

Are you sure the consumer can handle split lines? As vgacon_scrolldelta
(soff in particular) looks to me like it doesn't.

Have you tested you patch? I mean with soft scrollback on the vga console?

I only test the patch with the reproduce program.


Thanks,

Yang

quoted

  		vgacon_scrollback_cur->cnt++;
-		p += c->vc_size_row;
-		vgacon_scrollback_cur->tail += c->vc_size_row;
+		p += size;
+		vgacon_scrollback_cur->tail += size;
  
  		if (vgacon_scrollback_cur->tail >= vgacon_scrollback_cur->size)
  			vgacon_scrollback_cur->tail = 0;

thanks,