linux-kernel - fpga: fpga_mgr

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-next>] [day] [month] [year] [list]

Message-ID: <4617134.5euanDEBgJ@pcbe13614>
Date:   Thu, 21 Jun 2018 15:13:41 +0200
From:   Federico Vaga <federico.vaga@...n.ch>
To:     Alan Tull <atull@...nel.org>, Moritz Fischer <mdf@...nel.org>,
        <linux-fpga@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: fpga: fpga_mgr_get() buggy ?

Hello,

I believe that this patch

fpga: manager: change api, don't use drvdata
7085e2a94f7df5f419e3cfb2fe809ce6564e9629

is incomplete and buggy.

I completely agree that drvdata should not be used by the FPGA manager 
or any other subsystem like that.

What is buggy is the function fpga_mgr_get().
That patch has been done to allow multiple FPGA manager instances to 
be linked to the same device (PCI it says). But function 
fpga_mgr_get() will return only the first found: what about the 
others?

Then, all load kernel-doc comments says:

"This code assumes the caller got the mgr pointer from 
of_fpga_mgr_get() or fpga_mgr_get()"

but that function does not allow me to get, for instance, the second 
FPGA manager on my card.

Since, thanks to this patch I'm actually the creator of the 
fpga_manager structure,  I do not need to use fpga_mgr_get() to 
retrieve that data structure.
Despite this, I believe we still need to increment the module 
reference counter (which is done by fpga_mgr_get()).

We can fix this function by just replacing the argument from 'device' 
to 'fpga_manager' (the one returned by create() ). Alternatively, we 
can add an 'owner' field in "struct fpga_manager_ops" and 'get' it 
when we use it. Or again, just an 'owner' argument in the create() 
function. I'm proposing these alternatives because I'm not sure that 
this is correct:

	if (!try_module_get(dev->parent->driver->owner))

What if the device does not have a driver? Do we consider the 
following a valid use case?

probe(struct device *dev) {
  struct device *mydev;

  mydev->parent = dev;
  device_register(mydev);
  fpga_mrg_create(mydev, ....);
}

thanks :)