Hi all,

 

When I run linux-4.19.2 on arm64, I got panic in sys_imageblit. My test steps are as fllows:

1) Remote login using BMC;

2) Enter a bunch of 'Enter' keys;

3) Execute the 'top' command and continuously press shift + page up several times.

 

This gives the following panic on the serial console:

 

[47229.923256] Unable to handle kernel paging request at virtual address ffff00002f0bc010

[47229.923258] Mem abort info:

[47229.923259]   ESR = 0x96000047

[47229.923261]   Exception class = DABT (current EL), IL = 32 bits

[47229.923262]   SET = 0, FnV = 0

[47229.923263]   EA = 0, S1PTW = 0

[47229.923263] Data abort info:

[47229.923265]   ISV = 0, ISS = 0x00000047

[47229.923265]   CM = 0, WnR = 1

[47229.923269] swapper pgtable: 4k pages, 48-bit VAs, pgdp = 00000000ea9a862f

[47229.923270] [ffff00002f0bc010] pgd=000000dffbffe803, pud=000000dffbffd803, pmd=000000bf758ee003, pte=0000000000000000

[47229.923279] Internal error: Oops: 96000047 [#1] SMP

[47229.923285] CPU: 45 PID: 16061 Comm: top Kdump: loaded Tainted: G           OE     4.19.2-1.1.23.aarch64 #1

[47229.923286] Hardware name: Huawei Taishan 2280 /BC11SPCD, BIOS 1.46 03/29/2018

[47229.923288] pstate: 80000005 (Nzcv daif -PAN -UAO)

[47229.923295] pc : sys_imageblit+0x414/0x1000 [sysimgblt]

[47229.923317] lr : drm_fb_helper_sys_imageblit+0x28/0x50 [drm_kms_helper]

[47229.923318] sp : ffff00002f6d3820

[47229.923319] x29: ffff00002f6d3820 x28: 00000000000000ff

[47229.923322] x27: 0000000000000010 x26: 0000000000000001

[47229.923325] x25: 0000000000000118 x24: 0000000000000008

[47229.923327] x23: ffff805f771db800 x22: 0000000000000000

[47229.923330] x21: ffff00002f6d3958 x20: ffff00002f0bc010

[47229.923332] x19: 0000000000000000 x18: 0000000000000000

[47229.923335] x17: 0000000000000000 x16: 0000000000000000

[47229.923337] x15: 0000000000000000 x14: 2020202020202020

[47229.923339] x13: 2020202064686374 x12: ffff805f77927360

[47229.923342] x11: ffff000000f54078 x10: 0000000000000023

[47229.923344] x9 : 000000000000000f x8 : 0000000000000003

[47229.923346] x7 : 000000000000008c x6 : 0000000000000002

[47229.923349] x5 : 0000000000000000 x4 : 00000000ad55ad55

[47229.923352] x3 : ffff805f77927360 x2 : 0000000000000000

[47229.923354] x1 : 0000000000000006 x0 : 0000000000000000

[47229.923357] Process top (pid: 16061, stack limit = 0x00000000d490d156)

[47229.923359] Call trace:

[47229.923363]  sys_imageblit+0x414/0x1000 [sysimgblt]

[47229.923373]  drm_fb_helper_sys_imageblit+0x28/0x50 [drm_kms_helper]

[47229.923387]  bit_putcs+0x29c/0x4a0

[47229.923391]  fbcon_putcs+0x110/0x148

[47229.923397]  do_update_region+0x138/0x1d8

[47229.923399]  do_con_trol+0xb44/0x13e0

[47229.923400]  do_con_write.part.29+0x1d0/0x8f0

[47229.923402]  con_write+0x70/0x78

[47229.923405]  n_tty_write+0x1a8/0x428

[47229.923407]  tty_write+0x1bc/0x2f0

[47229.923412]  __vfs_write+0x60/0x1a8

[47229.923414]  vfs_write+0xb0/0x1a8

[47229.923416]  ksys_write+0x6c/0xd8

[47229.923418]  __arm64_sys_write+0x28/0x38

[47229.923424]  el0_svc_common+0xb8/0x118

[47229.923426]  el0_svc_handler+0x38/0x88

[47229.923429]  el0_svc+0x8/0xc

[47229.923431] Code: 0a080000 b8605960 0a000080 4a050000 (b8227a80)

[47229.923436] [kbox] die event detected

 

I followed the same test steps and found the same problem in the linux-4.20.0-rc3.

In the test I found that if the variable bitstart in function sys_imageblit is greater than variable p->screen_size - 1, the kernel will panic.

The following patch seems to work fine in my test:

 

diff --git a/drivers/video/fbdev/core/sysimgblt.c b/drivers/video/fbdev/core/sysimgblt.c

index a4d05b1..b316404 100644

--- a/drivers/video/fbdev/core/sysimgblt.c

+++ b/drivers/video/fbdev/core/sysimgblt.c

@@ -254,6 +254,9 @@ void sys_imageblit(struct fb_info *p, const struct fb_image *image)

        bitstart /= 8;

        bitstart &= ~(bpl - 1);

+ if (p->screen_size - 1 < bitstart)

+         return;

+

        dst1 = (void __force *)p->screen_base + bitstart;

        if (p->fbops->fb_sync)

 

So please CC me directly on any reply.

 

Many thanks,

 

Zengruan.