In __qedf_probe, if `qedf->cdev` is NULL which means
qed_ops->common->probe() failed, then the program will goto label err1,
scsi_host_put() will free `lport->host` pointer. Because the memory `qedf`
points to is allocated by libfc_host_alloc(), it will be freed by
scsi_host_put(). However, the if statement below label err0 only checks
whether qedf is NULL but doesn't check whether the memory has been freed.
So a UAF bug occurred.
There are two ways to get to the statements below err0. the first one is
described as before. `qedf` should be set to NULL. The second way is goto
`err0` directly. In this way `qedf` hasn't been changed and it has the
initial value `NULL`. So the program wont't go to the if statement in any
situation.
The KASAN logs are as follows:
[ 2.312969] BUG: KASAN: use-after-free in __qedf_probe+0x5dcf/0x6bc0
[ 2.312969]
[ 2.312969] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
[ 2.312969] Call Trace:
[ 2.312969] dump_stack_lvl+0x59/0x7b
[ 2.312969] print_address_description+0x7c/0x3b0
[ 2.312969] ? __qedf_probe+0x5dcf/0x6bc0
[ 2.312969] __kasan_report+0x160/0x1c0
[ 2.312969] ? __qedf_probe+0x5dcf/0x6bc0
[ 2.312969] kasan_report+0x4b/0x70
[ 2.312969] ? kobject_put+0x25d/0x290
[ 2.312969] kasan_check_range+0x2ca/0x310
[ 2.312969] __qedf_probe+0x5dcf/0x6bc0
[ 2.312969] ? selinux_kernfs_init_security+0xdc/0x5f0
[ 2.312969] ? trace_rpm_return_int_rcuidle+0x18/0x120
[ 2.312969] ? rpm_resume+0xa5c/0x16e0
[ 2.312969] ? qedf_get_generic_tlv_data+0x160/0x160
[ 2.312969] local_pci_probe+0x13c/0x1f0
[ 2.312969] pci_device_probe+0x37e/0x6c0
Reported-by: Zheyu Ma <[email protected]>
Signed-off-by: Letu Ren <[email protected]>
Co-developed-by: Wende Tan <[email protected]>
Signed-off-by: Wende Tan <[email protected]>
---
drivers/scsi/qedf/qedf_main.c | 5 -----
1 file changed, 5 deletions(-)
diff --git a/drivers/scsi/qedf/qedf_main.c b/drivers/scsi/qedf/qedf_main.c
index 42d0d941dba5..3ea552acd82a 100644
--- a/drivers/scsi/qedf/qedf_main.c
+++ b/drivers/scsi/qedf/qedf_main.c
@@ -3683,11 +3683,6 @@ static int __qedf_probe(struct pci_dev *pdev, int mode)
err1:
scsi_host_put(lport->host);
err0:
- if (qedf) {
- QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_DISC, "Probe done.\n");
-
- clear_bit(QEDF_PROBING, &qedf->flags);
- }
return rc;
}
--
2.33.1
Hi Letu Ren,
> -----Original Message-----
> From: Letu Ren <[email protected]>
> Sent: Friday, November 12, 2021 5:37 PM
> To: Saurav Kashyap <[email protected]>; Javed Hasan
> <[email protected]>; [email protected]; [email protected]
> Cc: GR-QLogic-Storage-Upstream <GR-QLogic-Storage-
> [email protected]>; [email protected]; linux-
> [email protected]; Letu Ren <[email protected]>; Zheyu Ma
> <[email protected]>; Wende Tan <[email protected]>
> Subject: [EXT] [PATCH] scsi: qedf: Fix a UAF bug in __qedf_probe
>
> External Email
>
> ----------------------------------------------------------------------
> In __qedf_probe, if `qedf->cdev` is NULL which means
> qed_ops->common->probe() failed, then the program will goto label err1,
> scsi_host_put() will free `lport->host` pointer. Because the memory `qedf`
> points to is allocated by libfc_host_alloc(), it will be freed by
> scsi_host_put(). However, the if statement below label err0 only checks
> whether qedf is NULL but doesn't check whether the memory has been freed.
> So a UAF bug occurred.
>
> There are two ways to get to the statements below err0. the first one is
> described as before. `qedf` should be set to NULL. The second way is goto
> `err0` directly. In this way `qedf` hasn't been changed and it has the
> initial value `NULL`. So the program wont't go to the if statement in any
> situation.
>
> The KASAN logs are as follows:
>
> [ 2.312969] BUG: KASAN: use-after-free in __qedf_probe+0x5dcf/0x6bc0
> [ 2.312969]
> [ 2.312969] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS
> rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
> [ 2.312969] Call Trace:
> [ 2.312969] dump_stack_lvl+0x59/0x7b
> [ 2.312969] print_address_description+0x7c/0x3b0
> [ 2.312969] ? __qedf_probe+0x5dcf/0x6bc0
> [ 2.312969] __kasan_report+0x160/0x1c0
> [ 2.312969] ? __qedf_probe+0x5dcf/0x6bc0
> [ 2.312969] kasan_report+0x4b/0x70
> [ 2.312969] ? kobject_put+0x25d/0x290
> [ 2.312969] kasan_check_range+0x2ca/0x310
> [ 2.312969] __qedf_probe+0x5dcf/0x6bc0
> [ 2.312969] ? selinux_kernfs_init_security+0xdc/0x5f0
> [ 2.312969] ? trace_rpm_return_int_rcuidle+0x18/0x120
> [ 2.312969] ? rpm_resume+0xa5c/0x16e0
> [ 2.312969] ? qedf_get_generic_tlv_data+0x160/0x160
> [ 2.312969] local_pci_probe+0x13c/0x1f0
> [ 2.312969] pci_device_probe+0x37e/0x6c0
>
> Reported-by: Zheyu Ma <[email protected]>
> Signed-off-by: Letu Ren <[email protected]>
> Co-developed-by: Wende Tan <[email protected]>
> Signed-off-by: Wende Tan <[email protected]>
> ---
> drivers/scsi/qedf/qedf_main.c | 5 -----
> 1 file changed, 5 deletions(-)
>
> diff --git a/drivers/scsi/qedf/qedf_main.c b/drivers/scsi/qedf/qedf_main.c
> index 42d0d941dba5..3ea552acd82a 100644
> --- a/drivers/scsi/qedf/qedf_main.c
> +++ b/drivers/scsi/qedf/qedf_main.c
> @@ -3683,11 +3683,6 @@ static int __qedf_probe(struct pci_dev *pdev, int
> mode)
> err1:
> scsi_host_put(lport->host);
> err0:
> - if (qedf) {
> - QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_DISC, "Probe
> done.\n");
> -
> - clear_bit(QEDF_PROBING, &qedf->flags);
> - }
> return rc;
> }
Thanks for the patch.
Acked-by: Saurav Kashyap <[email protected]>
Thanks,
~Saurav
>
> --
> 2.33.1
Hi Saurav,
> Thanks for the patch.
>
> Acked-by: Saurav Kashyap <[email protected]>
This is a friendly reminder. I wonder the review progress of this
patch. Almost a year has passed since I submitted the patch. And we
have tested qedf on Linux 6.0-rc0. This UAF problem still exists. If
there's anything I can do to help with this issue please let me know.
I'm willing to help.
Letu Ren
Hi Letu Ren,
I have already ack-ed this patch, so nothing to be done from my side.
Hi Martin,
I have ack-ed this patch, kindly pull it.
Thanks,
~saurav
> -----Original Message-----
> From: Letu Ren <[email protected]>
> Sent: Tuesday, September 6, 2022 3:31 PM
> To: Saurav Kashyap <[email protected]>
> Cc: Javed Hasan <[email protected]>; [email protected];
> [email protected]; GR-QLogic-Storage-Upstream <GR-QLogic-
> [email protected]>; [email protected]; linux-
> [email protected]; Zheyu Ma <[email protected]>; Wende Tan
> <[email protected]>
> Subject: Re: [EXT] [PATCH] scsi: qedf: Fix a UAF bug in __qedf_probe
>
> Hi Saurav,
>
> > Thanks for the patch.
> >
> > Acked-by: Saurav Kashyap <[email protected]>
>
>
> This is a friendly reminder. I wonder the review progress of this
> patch. Almost a year has passed since I submitted the patch. And we
> have tested qedf on Linux 6.0-rc0. This UAF problem still exists. If
> there's anything I can do to help with this issue please let me know.
> I'm willing to help.
>
> Letu Ren
On Fri, 12 Nov 2021 20:06:41 +0800, Letu Ren wrote:
> In __qedf_probe, if `qedf->cdev` is NULL which means
> qed_ops->common->probe() failed, then the program will goto label err1,
> scsi_host_put() will free `lport->host` pointer. Because the memory `qedf`
> points to is allocated by libfc_host_alloc(), it will be freed by
> scsi_host_put(). However, the if statement below label err0 only checks
> whether qedf is NULL but doesn't check whether the memory has been freed.
> So a UAF bug occurred.
>
> [...]
Applied to 6.0/scsi-fixes, thanks!
[1/1] scsi: qedf: Fix a UAF bug in __qedf_probe
https://git.kernel.org/mkp/scsi/c/fbfe96869b78
--
Martin K. Petersen Oracle Linux Engineering