linux-kernel - Re: [patch -target tree] usb: gadget: f

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]
Date:	Wed, 09 Mar 2016 21:19:23 -0800
From:	"Nicholas A. Bellinger" <nab@...ux-iscsi.org>
To:	Andrzej Pietrasiewicz <andrzej.p@...sung.com>
Cc:	Felipe Balbi <balbi@...nel.org>,
	Dan Carpenter <dan.carpenter@...cle.com>,
	Christoph Hellwig <hch@....de>,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
	Bart Van Assche <bart.vanassche@...disk.com>,
	linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org,
	target-devel@...r.kernel.org
Subject: Re: [patch -target tree] usb: gadget: f_tcm: use after free

Hi Andrzej,

On Wed, 2016-03-09 at 13:53 +0100, Andrzej Pietrasiewicz wrote:
> Hi Nicholas,
> 
> W dniu 05.03.2016 o 08:26, Nicholas A. Bellinger pisze:
> > Hi Felipe + usb-gadget folks,
> >
> > On Wed, 2016-03-02 at 13:55 +0200, Felipe Balbi wrote:
> 
> <snip>
> 
> >
> > usb-gadget/tcm: Conversion to percpu_ida tag pre-allocation
> > http://www.spinics.net/lists/target-devel/msg11777.html
> >
> > usb-gadget/tcm: Convert to TARGET_SCF_ACK_KREF I/O krefs
> > http://www.spinics.net/lists/target-devel/msg11782.html
> >
> > Felipe, Sebastian, & Andrezj, would you be so kind to review and test
> > usb-gadget using target-pending/for-next code..?
> >
> >
> 
> Actually I have noticed a problem at
> 8b00965 "target: Convert demo-mode only drivers to target_alloc_session"
> 
> I get:
> 
> [ 1698.406304] configfs-gadget gadget: high-speed config #1: c
> [ 1698.410547] Using the BOT protocol
> [ 1698.414163] Missing nexus, ignoring command
> [ 1698.417944] bot_cmd_complete(309): -22
> 
> I think the third message is from f_tcm. It is because
> tpg->tpg_nexus is not set anymore, because the line
> 
> 	tpg->tpg_nexus = tv_nexus;
> 
> is removed by the commit mentioned above.
> 
> Restoring this line fixes the problem.

Thanks for testing!

Applying the following patch to re-add the missing assingment
as a proper alloc_session callback.

diff --git a/drivers/usb/gadget/function/f_tcm.c b/drivers/usb/gadget/function/f_tcm.c
index e352a31..348140c 100644
--- a/drivers/usb/gadget/function/f_tcm.c
+++ b/drivers/usb/gadget/function/f_tcm.c
@@ -1570,6 +1570,16 @@ out:
        return ret;
 }
 
+static int usbg_alloc_sess_cb(struct se_portal_group *se_tpg,
+                             struct se_session *se_sess, void *p)
+{
+       struct usbg_tpg *tpg = container_of(se_tpg,
+                               struct usbg_tpg, se_tpg);
+
+       tpg->tpg_nexus = p;
+       return 0;
+}
+
 static int tcm_usbg_make_nexus(struct usbg_tpg *tpg, char *name)
 {
        struct tcm_usbg_nexus *tv_nexus;
@@ -1591,7 +1601,7 @@ static int tcm_usbg_make_nexus(struct usbg_tpg *tpg, char *name)
        tv_nexus->tvn_se_sess = target_alloc_session(&tpg->se_tpg, 128,
                                                     sizeof(struct usbg_cmd),
                                                     TARGET_PROT_NORMAL, name,
-                                                    tv_nexus, NULL);
+                                                    tv_nexus, usbg_alloc_sess_cb);
        if (IS_ERR(tv_nexus->tvn_se_sess)) {
 #define MAKE_NEXUS_MSG "core_tpg_check_initiator_node_acl() failed for %s\n"
                pr_debug(MAKE_NEXUS_MSG, name);

> 
> Then percpu_ida commit produces the below result
> (it does not happen immediately, but a while after
> running the script).
> I did not bisect, though; I only checked the commits
> which alter drivers/usb/gadget/function/f_tcm.c.
> The last one which (almost) works is:
> 
> 8b00965 "target: Convert demo-mode only drivers to target_alloc_session"
> 
> AP
> 
> I used the following script:
> 
> 8<----------------------------
> 
> #!/bin/bash
> 
> mount -t configfs none /sys/kernel/config
> modprobe libcomposite
> cd /sys/kernel/config/usb_gadget
> mkdir tcm
> cd tcm
> mkdir functions/tcm.0
> cd /sys/kernel/config/target/
> mkdir usb_gadget
> cd usb_gadget
> mkdir naa.0123456789abcdef
> cd naa.0123456789abcdef
> mkdir tpgt_1
> cd tpgt_1
> echo naa.01234567890abcdef > nexus
> echo 1 > enable
> cd /sys/kernel/config/usb_gadget/tcm
> mkdir configs/c.1
> ln -s functions/tcm.0 configs/c.1
> echo 0x0525 > idVendor
> echo 0x1234 > idProduct
> echo 12400000.dwc3 > UDC
> 
> 8<----------------------------
> 
> # [   45.510513] ERR bot_status_complete(73)
> [   45.671921] configfs-gadget gadget: high-speed config #1: c
> [   45.676158] Using the BOT protocol
> [   45.679860] ------------[ cut here ]------------
> [   45.683981] kernel BUG at drivers/target/target_core_transport.c:1476!

Mmmm, usbg_get_cmd() was missing an explicit memset() after tag lookup.

How about the following..?

diff --git a/drivers/usb/gadget/function/f_tcm.c b/drivers/usb/gadget/function/f_tcm.c
index e352a31..d4e8a91 100644
--- a/drivers/usb/gadget/function/f_tcm.c
+++ b/drivers/usb/gadget/function/f_tcm.c
@@ -1078,6 +1078,7 @@ static struct usbg_cmd *usbg_get_cmd(struct f_uas *fu,
                return ERR_PTR(-ENOMEM);
 
        cmd = &((struct usbg_cmd *)se_sess->sess_cmd_map)[tag];
+       memset(cmd, 0, sizeof(*cmd));
        cmd->se_cmd.map_tag = tag;
        cmd->se_cmd.tag = cmd->tag = scsi_tag;
        cmd->fu = fu;