linux-kernel - Re: [PATCH rdma-next v1 02/15] RDMA/core: Replace the ib_port_data hw_stats pointers with a ib

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:   Mon, 7 Jun 2021 09:32:10 -0300
From:   Jason Gunthorpe <jgg@...dia.com>
To:     Greg KH <gregkh@...uxfoundation.org>
Cc:     Leon Romanovsky <leon@...nel.org>,
        Doug Ledford <dledford@...hat.com>,
        Kees Cook <keescook@...omium.org>,
        Nathan Chancellor <nathan@...nel.org>,
        Adit Ranadive <aditr@...are.com>,
        Ariel Elior <aelior@...vell.com>,
        Christian Benvenuti <benve@...co.com>,
        clang-built-linux@...glegroups.com,
        Dennis Dalessandro <dennis.dalessandro@...nelisnetworks.com>,
        Devesh Sharma <devesh.sharma@...adcom.com>,
        Gal Pressman <galpress@...zon.com>,
        linux-kernel@...r.kernel.org, linux-rdma@...r.kernel.org,
        Michal Kalderon <mkalderon@...vell.com>,
        Mike Marciniszyn <mike.marciniszyn@...nelisnetworks.com>,
        Mustafa Ismail <mustafa.ismail@...el.com>,
        Naresh Kumar PBS <nareshkumar.pbs@...adcom.com>,
        Nelson Escobar <neescoba@...co.com>,
        Nick Desaulniers <ndesaulniers@...gle.com>,
        Potnuri Bharat Teja <bharat@...lsio.com>,
        Selvin Xavier <selvin.xavier@...adcom.com>,
        Shiraz Saleem <shiraz.saleem@...el.com>,
        VMware PV-Drivers <pv-drivers@...are.com>,
        Yishai Hadas <yishaih@...dia.com>,
        Zhu Yanjun <zyjzyj2000@...il.com>
Subject: Re: [PATCH rdma-next v1 02/15] RDMA/core: Replace the ib_port_data
 hw_stats pointers with a ib_port pointer

On Mon, Jun 07, 2021 at 12:23:23PM +0200, Greg KH wrote:
> >  static int add_port(struct ib_core_device *coredev, int port_num)
> >  {
> >  	struct ib_device *device = rdma_device_to_ibdev(&coredev->dev);
> > @@ -1171,6 +1177,8 @@ static int add_port(struct ib_core_device *coredev, int port_num)
> >  		setup_hw_stats(device, p, port_num);
> >  
> >  	list_add_tail(&p->kobj.entry, &coredev->port_list);
> > +	if (device->port_data && is_full_dev)
> > +		device->port_data[port_num].sysfs = p;
> 
> You are saving off a pointer to a reference counted structure without
> incrementing the reference count on it?  

This storage borrows another reference count, primarily because there
is no locking to read/write .sysfs. It is a fairly common idiom.

You can see it in the free path:

		port->ibdev->port_data[port->port_num].sysfs = NULL;
	kobject_put(&port->kobj);  // port == p above

Due to the lack of locks the whole external thing is arranged so that
the 3 users of .sysfs are sequenced properly around
setup_port()/destroy_port() using other external locks.

Adding more refs without also adding locking is just confusing what
the data protection model is. This is a borrowed ref and access is
only allowed when other locking is properly sequencing it with the ref
owner's manipulation of .sysfs.

Eg I would reject some code sequence like this:

	port->ibdev->port_data[port->port_num].sysfs = NULL;
	kobject_put(&port->kobj);  // one for .sysfs
	kobject_put(&port->kobj);  // one for our stack

As being pretty bogus.

Jason