Skip to contentSkip to conversation

ubi: Reject device with erasesize 0

PATCH In review Jan Kara 4 participants 8 messages Aug 31, 2023 – Sep 1, 2023

Subsystems (2)
lore
Share & tools
lore

Jan Kara

author
#0 lore raw

PATCH 1/1 ubi: Reject device with erasesize 0

In principle MTD device with erasesize 0 can exist and it is possible to create them e.g. via KVM. If that happens UBI layer currently crashes with:

ubi7: attaching mtd147
divide error: 0000 [#1] PREEMPT SMP KASAN

CPU: 0 PID: 20023 Comm: syz-executor.0 Not tainted 6.2.0 #6
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
1.13.0-1ubuntu1.1 04/01/2014
RIP: 0010:mtd_div_by_eb include/linux/mtd/mtd.h:580 [inline]
RIP: 0010:io_init drivers/mtd/ubi/build.c:620 [inline]
RIP: 0010:ubi_attach_mtd_dev+0x77f/0x2fe0 drivers/mtd/ubi/build.c:955
Code: fc ff df 48 c1 ea 03 0f b6 14 02 4c 89 f0 83 e0 07 83 c0 03 38
d0 7c 08 84 d2 0f 85 1f 25 00 00 41 8b 4c 24 10 48 89 d8 31 d2 <48> f7
f1 48 89 c3 e8 b6 f3 1b fc 48 8d 85 40 17 00 00 48 89 c2 48
RSP: 0018:ffffc9000be0fd30 EFLAGS: 00010246
RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffff888047a49d40 RDI: 0000000000000002
RBP: ffff888024e1c000 R08: 0000000000000016 R09: fffff520017c1f47
R10: ffffc9000be0fa37 R11: fffff520017c1f46 R12: ffff88806545a000
R13: 0000000000000000 R14: ffff88806545a010 R15: 0000000000000007
FS:  00007fd45e85c700(0000) GS:ffff88802ca00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f64aeef53a4 CR3: 000000004f39a000 CR4: 0000000000350ef0
Call Trace:
 <TASK>
 ctrl_cdev_ioctl+0x303/0x3a0 drivers/mtd/ubi/cdev.c:1043
 vfs_ioctl fs/ioctl.c:51 [inline]
 __do_sys_ioctl fs/ioctl.c:870 [inline]
 __se_sys_ioctl fs/ioctl.c:856 [inline]
 __x64_sys_ioctl+0x198/0x210 fs/ioctl.c:856
 do_syscall_x64 arch/x86/entry/common.c:50 [inline]
 do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
 entry_SYSCALL_64_after_hwframe+0x63/0xcd

Handle such devices gracefully and just reject attaching UBI to them instead of crashing.

---
 drivers/mtd/ubi/build.c | 3 +++
 1 file changed, 3 insertions(+)

It doesn't seem the discussion linked above concluded in a patch. So is anything wrong with the trivial approach here so that we can close the issue?

drivers/mtd/ubi/build.c +3 -0
index 8b91a55ec0d2..684273e13efb 100644
@@ -613,6 +613,9 @@ static int io_init(struct ubi_device *ubi, int max_beb_per1024)
 	if (ubi->vid_hdr_offset < 0)
 		return -EINVAL;
 
+	if (ubi->mtd->erasesize == 0)
+		return -EINVAL;
+
 	/*
 	 * Note, in this implementation we support MTD devices with 0x7FFFFFFF
 	 * physical eraseblocks maximum.
Signature · 6 lines
2.35.3


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
reviewer
#1 lore raw

在 2023/8/31 19:11, Jan Kara 写道:
Hi Jan,

Quoted 41 lines

In principle MTD device with erasesize 0 can exist and it is possible to create them e.g. via KVM. If that happens UBI layer currently crashes with:

ubi7: attaching mtd147
divide error: 0000 [#1] PREEMPT SMP KASAN

CPU: 0 PID: 20023 Comm: syz-executor.0 Not tainted 6.2.0 #6
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
1.13.0-1ubuntu1.1 04/01/2014
RIP: 0010:mtd_div_by_eb include/linux/mtd/mtd.h:580 [inline]
RIP: 0010:io_init drivers/mtd/ubi/build.c:620 [inline]
RIP: 0010:ubi_attach_mtd_dev+0x77f/0x2fe0 drivers/mtd/ubi/build.c:955
Code: fc ff df 48 c1 ea 03 0f b6 14 02 4c 89 f0 83 e0 07 83 c0 03 38
d0 7c 08 84 d2 0f 85 1f 25 00 00 41 8b 4c 24 10 48 89 d8 31 d2 <48> f7
f1 48 89 c3 e8 b6 f3 1b fc 48 8d 85 40 17 00 00 48 89 c2 48
RSP: 0018:ffffc9000be0fd30 EFLAGS: 00010246
RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffff888047a49d40 RDI: 0000000000000002
RBP: ffff888024e1c000 R08: 0000000000000016 R09: fffff520017c1f47
R10: ffffc9000be0fa37 R11: fffff520017c1f46 R12: ffff88806545a000
R13: 0000000000000000 R14: ffff88806545a010 R15: 0000000000000007
FS:  00007fd45e85c700(0000) GS:ffff88802ca00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f64aeef53a4 CR3: 000000004f39a000 CR4: 0000000000350ef0
Call Trace:
  <TASK>
  ctrl_cdev_ioctl+0x303/0x3a0 drivers/mtd/ubi/cdev.c:1043
  vfs_ioctl fs/ioctl.c:51 [inline]
  __do_sys_ioctl fs/ioctl.c:870 [inline]
  __se_sys_ioctl fs/ioctl.c:856 [inline]
  __x64_sys_ioctl+0x198/0x210 fs/ioctl.c:856
  do_syscall_x64 arch/x86/entry/common.c:50 [inline]
  do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
  entry_SYSCALL_64_after_hwframe+0x63/0xcd

Handle such devices gracefully and just reject attaching UBI to them instead of crashing.

Quoted 6 lines
---
  drivers/mtd/ubi/build.c | 3 +++
  1 file changed, 3 insertions(+)

It doesn't seem the discussion linked above concluded in a patch. So is anything wrong with the trivial approach here so that we can close the issue?

There was a patch, see https://patchwork.ozlabs.org/project/linux-mtd/patch/[email protected]/

Quoted 7 lines
drivers/mtd/ubi/build.c +0 -0
index 8b91a55ec0d2..684273e13efb 100644

drivers/mtd/ubi/build.c static int io_init(struct ubi_device *ubi, int max_beb_per1024) · line 621

Quoted 9 lines
drivers/mtd/ubi/build.c +3 -0
@@ -613,6 +613,9 @@ static int io_init(struct ubi_device *ubi, int max_beb_per1024)
  	if (ubi->vid_hdr_offset < 0)
  		return -EINVAL;
 
+	if (ubi->mtd->erasesize == 0)
+		return -EINVAL;
+
  	/*
  	 * Note, in this implementation we support MTD devices with 0x7FFFFFFF
  	 * physical eraseblocks maximum.

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

#2 lore raw

----- Ursprüngliche Mail -----

Quoted 7 lines

Von: "chengzhihao1" <[email protected]>

Quoted 6 lines

https://lore.kernel.org/all/CA+UBctDsHRpkLG5ppdiuV8Msn4Dx-ZJ2xDrxfa48VMb7ZE+xBA@mail.gmail.com

---
  drivers/mtd/ubi/build.c | 3 +++
  1 file changed, 3 insertions(+)
Quoted 5 lines
Quoted 2 lines

It doesn't seem the discussion linked above concluded in a patch. So is anything wrong with the trivial approach here so that we can close the issue?

There was a patch, see
https://patchwork.ozlabs.org/project/linux-mtd/patch/[email protected]/

Hmm, let me check the patch status of that one.

Thanks,
//richard

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

author
#3 lore raw
In reply to Zhihao Cheng
Quoted 47 lines · Zhihao Cheng

在 2023/8/31 19:11, Jan Kara 写道:
Hi Jan,

Quoted 45 lines

In principle MTD device with erasesize 0 can exist and it is possible to create them e.g. via KVM. If that happens UBI layer currently crashes with:

ubi7: attaching mtd147
divide error: 0000 [#1] PREEMPT SMP KASAN

CPU: 0 PID: 20023 Comm: syz-executor.0 Not tainted 6.2.0 #6
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
1.13.0-1ubuntu1.1 04/01/2014
RIP: 0010:mtd_div_by_eb include/linux/mtd/mtd.h:580 [inline]
RIP: 0010:io_init drivers/mtd/ubi/build.c:620 [inline]
RIP: 0010:ubi_attach_mtd_dev+0x77f/0x2fe0 drivers/mtd/ubi/build.c:955
Code: fc ff df 48 c1 ea 03 0f b6 14 02 4c 89 f0 83 e0 07 83 c0 03 38
d0 7c 08 84 d2 0f 85 1f 25 00 00 41 8b 4c 24 10 48 89 d8 31 d2 <48> f7
f1 48 89 c3 e8 b6 f3 1b fc 48 8d 85 40 17 00 00 48 89 c2 48
RSP: 0018:ffffc9000be0fd30 EFLAGS: 00010246
RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffff888047a49d40 RDI: 0000000000000002
RBP: ffff888024e1c000 R08: 0000000000000016 R09: fffff520017c1f47
R10: ffffc9000be0fa37 R11: fffff520017c1f46 R12: ffff88806545a000
R13: 0000000000000000 R14: ffff88806545a010 R15: 0000000000000007
FS:  00007fd45e85c700(0000) GS:ffff88802ca00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f64aeef53a4 CR3: 000000004f39a000 CR4: 0000000000350ef0
Call Trace:
  <TASK>
  ctrl_cdev_ioctl+0x303/0x3a0 drivers/mtd/ubi/cdev.c:1043
  vfs_ioctl fs/ioctl.c:51 [inline]
  __do_sys_ioctl fs/ioctl.c:870 [inline]
  __se_sys_ioctl fs/ioctl.c:856 [inline]
  __x64_sys_ioctl+0x198/0x210 fs/ioctl.c:856
  do_syscall_x64 arch/x86/entry/common.c:50 [inline]
  do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
  entry_SYSCALL_64_after_hwframe+0x63/0xcd

Handle such devices gracefully and just reject attaching UBI to them instead of crashing.

---
  drivers/mtd/ubi/build.c | 3 +++
  1 file changed, 3 insertions(+)
Quoted 5 lines
Quoted 2 lines

It doesn't seem the discussion linked above concluded in a patch. So is anything wrong with the trivial approach here so that we can close the issue?

There was a patch, see
https://patchwork.ozlabs.org/project/linux-mtd/patch/[email protected]/

Ah, OK, I wasn't able to find it through lore search. Thanks for the pointer! Hopefully Richard will merge it now.

Honza

Signature · 6 lines
Jan Kara <[email protected]>
SUSE Labs, CR

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
#4 lore raw
In reply to Jan Kara

Hi, Richard,

Quoted 1 line · Jan Kara

In principle MTD device with erasesize 0 can exist and it is possible to create them e.g. via KVM. If that happens UBI layer currently crashes

mtd devices are organized in erase blocks, having a mtd with erase size zero looks suspicious to me. Do you know if there are any mtd devices with erasesize of zero?

Cheers,
ta

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

author
#5 lore raw
Quoted 8 lines · Tudor Ambarus

Hi, Richard,

Quoted 1 line · Jan Kara

In principle MTD device with erasesize 0 can exist and it is possible to create them e.g. via KVM. If that happens UBI layer currently crashes

mtd devices are organized in erase blocks, having a mtd with erase size zero looks suspicious to me. Do you know if there are any mtd devices with erasesize of zero?

Please see the discussion I've referenced from the patch. I agree such devices are strange but apparently devices reporting erasesize==0 do exist and can be created in virtualized environment. IMO we don't need to work with them but we shouldn't crash a kernel if they are attached...

Honza

Signature · 6 lines
Jan Kara <[email protected]>
SUSE Labs, CR

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
#6 lore raw
Quoted 10 lines · Jan Kara
Quoted 8 lines · Tudor Ambarus

Hi, Richard,

Quoted 1 line · Jan Kara

In principle MTD device with erasesize 0 can exist and it is possible to create them e.g. via KVM. If that happens UBI layer currently crashes

mtd devices are organized in erase blocks, having a mtd with erase size zero looks suspicious to me. Do you know if there are any mtd devices with erasesize of zero?

Quoted 7 lines

Please see the discussion I've referenced from the patch. I agree such devices are strange but apparently devices reporting erasesize==0 do exist and can be created in virtualized environment. IMO we don't need to work with them but we shouldn't crash a kernel if they are attached...

Honza

Thanks, Jan. I agree we shouldn't crash the kernel. I was wondering whether on the long term we should aim to not allow the creation of mtd devices with erasesize zero, or/and to not accept devices with erasesize of zero under mtd. Was curious about the why, I don't have anything against the patch. I guess I'll answer my questions myself if I ever stumble on this.

Cheers,
ta

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

#7 lore raw
In reply to Tudor Ambarus

----- Ursprüngliche Mail -----

Quoted 8 lines

Von: "Tudor Ambarus" <[email protected]>
An: "Jan Kara" <[email protected]>, "richard" <[email protected]>, "Miquel Raynal" <[email protected]>
CC: "linux-mtd" <[email protected]>, "chengzhihao1" <[email protected]>, "Yu Hao" <[email protected]>
Gesendet: Freitag, 1. September 2023 11:25:11
Betreff: Re: [PATCH] ubi: Reject device with erasesize 0

Hi, Richard,

Quoted 7 lines
Quoted 2 lines · Jan Kara

In principle MTD device with erasesize 0 can exist and it is possible to create them e.g. via KVM. If that happens UBI layer currently crashes

mtd devices are organized in erase blocks, having a mtd with erase size zero looks suspicious to me. Do you know if there are any mtd devices with erasesize of zero?

Yes. There are. So, the fix is fine.

Thanks,
//richard

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/