2020-01-14 09:36:07

by quanyang wang

Subject: [PATCH] ubi: fix memory leak from ubi->fm_anchor

From: Quanyang Wang <[email protected]>

Some ubi_wl_entry are allocated in erase_aeb() and one of them is
assigned to ubi->fm_anchor in __erase_worker(). And it should be freed
like others which are freed in tree_destroy(). Otherwise, it will
cause a memory leak:

unreferenced object 0xbc094318 (size 24):
comm "ubiattach", pid 491, jiffies 4294954015 (age 420.110s)
hex dump (first 24 bytes):
30 43 09 bc 00 00 00 00 00 00 00 00 01 00 00 00 0C..............
02 00 00 00 04 00 00 00 ........
backtrace:
[<6c2d5089>] erase_aeb+0x28/0xc8
[<a1c68fb1>] ubi_wl_init+0x1d8/0x4a8
[<d4f408f8>] ubi_attach+0xffc/0x10d0
[<add3b5d8>] ubi_attach_mtd_dev+0x5b4/0x9fc
[<d375a11c>] ctrl_cdev_ioctl+0xb8/0x1d8
[<72b250f2>] vfs_ioctl+0x28/0x3c
[<b80095d7>] do_vfs_ioctl+0xb0/0x798
[<bf9ef69e>] ksys_ioctl+0x58/0x74
[<5355bdbe>] ret_fast_syscall+0x0/0x54
[<90c6c3ca>] 0x7eadf854

Signed-off-by: Quanyang Wang <[email protected]>
---
drivers/mtd/ubi/wl.c | 2 ++
1 file changed, 2 insertions(+)

diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
index 5d77a38dba54..a5e9d1e4dc34 100644
--- a/drivers/mtd/ubi/wl.c
+++ b/drivers/mtd/ubi/wl.c
@@ -1885,6 +1885,7 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
tree_destroy(ubi, &ubi->used);
tree_destroy(ubi, &ubi->free);
tree_destroy(ubi, &ubi->scrub);
+ wl_entry_destroy(ubi, ubi->fm_anchor);
kfree(ubi->lookuptbl);
return err;
}
@@ -1920,6 +1921,7 @@ void ubi_wl_close(struct ubi_device *ubi)
tree_destroy(ubi, &ubi->erroneous);
tree_destroy(ubi, &ubi->free);
tree_destroy(ubi, &ubi->scrub);
+ wl_entry_destroy(ubi, ubi->fm_anchor);
kfree(ubi->lookuptbl);
}

--
2.17.1

2020-02-03 10:01:20

by quanyang wang

Subject: Re: [PATCH] ubi: fix memory leak from ubi->fm_anchor

Ping?

On 1/14/20 5:33 PM, [email protected] wrote:
> From: Quanyang Wang <[email protected]>
>
> Some ubi_wl_entry are allocated in erase_aeb() and one of them is
> assigned to ubi->fm_anchor in __erase_worker(). And it should be freed
> like others which are freed in tree_destroy(). Otherwise, it will
> cause a memory leak:
>
> unreferenced object 0xbc094318 (size 24):
> comm "ubiattach", pid 491, jiffies 4294954015 (age 420.110s)
> hex dump (first 24 bytes):
> 30 43 09 bc 00 00 00 00 00 00 00 00 01 00 00 00 0C..............
> 02 00 00 00 04 00 00 00 ........
> backtrace:
> [<6c2d5089>] erase_aeb+0x28/0xc8
> [<a1c68fb1>] ubi_wl_init+0x1d8/0x4a8
> [<d4f408f8>] ubi_attach+0xffc/0x10d0
> [<add3b5d8>] ubi_attach_mtd_dev+0x5b4/0x9fc
> [<d375a11c>] ctrl_cdev_ioctl+0xb8/0x1d8
> [<72b250f2>] vfs_ioctl+0x28/0x3c
> [<b80095d7>] do_vfs_ioctl+0xb0/0x798
> [<bf9ef69e>] ksys_ioctl+0x58/0x74
> [<5355bdbe>] ret_fast_syscall+0x0/0x54
> [<90c6c3ca>] 0x7eadf854
>
> Signed-off-by: Quanyang Wang <[email protected]>
> ---
> drivers/mtd/ubi/wl.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
> index 5d77a38dba54..a5e9d1e4dc34 100644
> --- a/drivers/mtd/ubi/wl.c
> +++ b/drivers/mtd/ubi/wl.c
> @@ -1885,6 +1885,7 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
> tree_destroy(ubi, &ubi->used);
> tree_destroy(ubi, &ubi->free);
> tree_destroy(ubi, &ubi->scrub);
> + wl_entry_destroy(ubi, ubi->fm_anchor);
> kfree(ubi->lookuptbl);
> return err;
> }
> @@ -1920,6 +1921,7 @@ void ubi_wl_close(struct ubi_device *ubi)
> tree_destroy(ubi, &ubi->erroneous);
> tree_destroy(ubi, &ubi->free);
> tree_destroy(ubi, &ubi->scrub);
> + wl_entry_destroy(ubi, ubi->fm_anchor);
> kfree(ubi->lookuptbl);
> }
>

2020-02-07 15:55:44

by Richard Weinberger

Subject: Re: [PATCH] ubi: fix memory leak from ubi->fm_anchor

On Mon, Feb 3, 2020 at 10:14 AM Quanyang Wang
<[email protected]> wrote:
>
> Ping?
>
> On 1/14/20 5:33 PM, [email protected] wrote:
> > From: Quanyang Wang <[email protected]>
> >
> > Some ubi_wl_entry are allocated in erase_aeb() and one of them is
> > assigned to ubi->fm_anchor in __erase_worker(). And it should be freed
> > like others which are freed in tree_destroy(). Otherwise, it will
> > cause a memory leak:
> >
> > unreferenced object 0xbc094318 (size 24):
> > comm "ubiattach", pid 491, jiffies 4294954015 (age 420.110s)
> > hex dump (first 24 bytes):
> > 30 43 09 bc 00 00 00 00 00 00 00 00 01 00 00 00 0C..............
> > 02 00 00 00 04 00 00 00 ........
> > backtrace:
> > [<6c2d5089>] erase_aeb+0x28/0xc8
> > [<a1c68fb1>] ubi_wl_init+0x1d8/0x4a8
> > [<d4f408f8>] ubi_attach+0xffc/0x10d0
> > [<add3b5d8>] ubi_attach_mtd_dev+0x5b4/0x9fc
> > [<d375a11c>] ctrl_cdev_ioctl+0xb8/0x1d8
> > [<72b250f2>] vfs_ioctl+0x28/0x3c
> > [<b80095d7>] do_vfs_ioctl+0xb0/0x798
> > [<bf9ef69e>] ksys_ioctl+0x58/0x74
> > [<5355bdbe>] ret_fast_syscall+0x0/0x54
> > [<90c6c3ca>] 0x7eadf854
> >
> > Signed-off-by: Quanyang Wang <[email protected]>
> > ---
> > drivers/mtd/ubi/wl.c | 2 ++
> > 1 file changed, 2 insertions(+)

Good catch!
Fixes: f9c34bb52997 ("ubi: Fix producing anchor PEBs")

---
Thanks,
//richard

2020-02-08 03:10:05

by Hou Tao

Subject: Re: [PATCH] ubi: fix memory leak from ubi->fm_anchor

Hi,

The same problem has already been fixed by the patch in the following link early:

https://lore.kernel.org/linux-mtd/[email protected]/T/#medffabe29b65eb5feb387bff84c6ec7ad235c310

I will send a v2 next week.

Regards,
Tao

On 2020/2/7 23:54, Richard Weinberger wrote:
> On Mon, Feb 3, 2020 at 10:14 AM Quanyang Wang
> <[email protected]> wrote:
>>
>> Ping?
>>
>> On 1/14/20 5:33 PM, [email protected] wrote:
>>> From: Quanyang Wang <[email protected]>
>>>
>>> Some ubi_wl_entry are allocated in erase_aeb() and one of them is
>>> assigned to ubi->fm_anchor in __erase_worker(). And it should be freed
>>> like others which are freed in tree_destroy(). Otherwise, it will
>>> cause a memory leak:
>>>
>>> unreferenced object 0xbc094318 (size 24):
>>> comm "ubiattach", pid 491, jiffies 4294954015 (age 420.110s)
>>> hex dump (first 24 bytes):
>>> 30 43 09 bc 00 00 00 00 00 00 00 00 01 00 00 00 0C..............
>>> 02 00 00 00 04 00 00 00 ........
>>> backtrace:
>>> [<6c2d5089>] erase_aeb+0x28/0xc8
>>> [<a1c68fb1>] ubi_wl_init+0x1d8/0x4a8
>>> [<d4f408f8>] ubi_attach+0xffc/0x10d0
>>> [<add3b5d8>] ubi_attach_mtd_dev+0x5b4/0x9fc
>>> [<d375a11c>] ctrl_cdev_ioctl+0xb8/0x1d8
>>> [<72b250f2>] vfs_ioctl+0x28/0x3c
>>> [<b80095d7>] do_vfs_ioctl+0xb0/0x798
>>> [<bf9ef69e>] ksys_ioctl+0x58/0x74
>>> [<5355bdbe>] ret_fast_syscall+0x0/0x54
>>> [<90c6c3ca>] 0x7eadf854
>>>
>>> Signed-off-by: Quanyang Wang <[email protected]>
>>> ---
>>> drivers/mtd/ubi/wl.c | 2 ++
>>> 1 file changed, 2 insertions(+)
>
> Good catch!
> Fixes: f9c34bb52997 ("ubi: Fix producing anchor PEBs")
>
> ---
> Thanks,
> //richard
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
>
>