linux-kernel - Re: [PATCH] scsi: be2iscsi: Fix a use after free in beiscsi_if_clr

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <yq11r8kibzu.fsf@ca-mkp.ca.oracle.com>
Date:   Tue, 29 Jun 2021 18:02:08 -0400
From:   "Martin K. Petersen" <martin.petersen@...cle.com>
To:     Lv Yunlong <lyl2019@...l.ustc.edu.cn>
Cc:     subbu.seetharaman@...adcom.com, ketan.mukadam@...adcom.com,
        jitendra.bhivare@...adcom.com, jejb@...ux.ibm.com,
        martin.petersen@...cle.com, linux-scsi@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Mike Christie <michael.christie@...cle.com>
Subject: Re: [PATCH] scsi: be2iscsi: Fix a use after free in beiscsi_if_clr_ip

Lv,

> In the free_cmd error path of callee beiscsi_exec_nemb_cmd(),
> nonemb_cmd->va is freed by dma_free_coherent().  As req =
> nonemb_cmd.va, we can see that the freed nonemb_cmd.va is still
> dereferenced and used by req->ip_params.ip_record.status.

> My patch uses status to replace req->ip_params.ip_record.status to
> avoid the uaf.

This status is captured prior to executing the command so it doesn't
actually reflect whether the operation was successful (which I believe
was the intent).

Some of the callers of beiscsi_exec_nemb_cmd() pass a response buffer
and a response length as the two last arguments. Since
beiscsi_exec_nemb_cmd() frees the command before returning, passing a
response buffer seems to be the only way to get meaningful data out.

I am not sure whether the OPCODE_COMMON_ISCSI_NTWK_MODIFY_IP_ADDR
operation returns something useful from the controller. As far as I can
tell not all operations have a response buffer defined.

My recommendation would be to add a response buffer and try to decipher
what comes back from the firmware. Also, beiscsi_if_set_ip() appears to
have the same problem as beiscsi_if_clr_ip().

-- 
Martin K. Petersen	Oracle Linux Engineering