From mboxrd@z Thu Jan 1 00:00:00 1970 From: Richard Fitzgerald Subject: Re: [PATCH net-next v2 4/7] net: phy: mscc: read 'vsc8531,edge-slowdown' as an u32 [UNSCANNED] Date: Tue, 4 Sep 2018 10:21:51 +0100 Message-ID: <6c189a8f-7168-3861-c964-b38631833451@opensource.cirrus.com> References: <20180903084853.18092-1-quentin.schulz@bootlin.com> <20180903084853.18092-4-quentin.schulz@bootlin.com> <20180903132756.GD4445@lunn.ch> <20180903133746.wsvezy3rbdivnjfs@qschulz> <20180903200554.GJ4445@lunn.ch> <20180904072630.zc6sdz2xdti5nku4@qschulz> Mime-Version: 1.0 Content-Type: text/plain; charset="windows-1252"; format=flowed Content-Transfer-Encoding: 8bit Cc: , , , , , , , , , To: Quentin Schulz , Andrew Lunn Return-path: In-Reply-To: <20180904072630.zc6sdz2xdti5nku4@qschulz> Content-Language: en-US Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On 04/09/18 08:26, Quentin Schulz wrote: > Hi Andrew, > > On Mon, Sep 03, 2018 at 10:05:54PM +0200, Andrew Lunn wrote: >>> Just to be sure, we're talking here about making sure the value stored >>> in the DT is not bigger than the specified value (here an u8)? If so, >>> that isn't the reason why I'm suggesting those two patches. >>> >>> Without /bits 8/ in the DT property, whatever were the values I put in >>> the property, I'd always get a 0. So I need to fix it either in the DT >>> (but Rob does not really like it) or in the driver. >> >> Hi Quentin >> >> Ah, you are fixing endian issues. That was not clear to me from the >> commit message. >> >> I don't know enough about how DT stores values in the blob. Is there >> type info? Can the DT core tell if a value in the blob is a u8 or a >> u32? It would be nice if it warned about reading a u8 from a u32 >> blob. >> > > From my quick research, the lower bound checking is performed by > of_property_read_u* functions but not the higher bound checking (the > internal function of_find_property_value_of_size allows higher bound > checking but it seems it's never used by those functions (see 0 in > sz_max of of_property_read_variable_u*_array)). > > sz_max is used by of_property_read_variable_u*_array to copy at a > maximum of sz_max values in the output buffer. If sz_max is 0, it takes > sz_min so it's an array of definite size. > So since sz_max is 0 for all calls to of_property_read_variable_u*_array > by of_property_read_u*_array, we basically know we'll get a buffer of > sz_min values but we don't actually make use of the higher bound > checking of of_find_property_value_of_size. > This was the original behaviour of the of_property_read_u*_array functions. If you look back at the of_property_read_u*_array implementations before my patch they passed max=0 to of_find_property_value_of_size. To avoid duplicating code I reimplemented the of_property_read_u*_array to use the new of_property_read_variable_u*_array hence they pass sz_max=0 to preserve the original behaviour that max=0 to of_find_property_value_of_size, so that I didn't break any code that might depend on that. > We could enforce this higher bound check by, instead of setting sz_max > to 0, setting sz_max to sz_min in calls to of_property_read_u*_array. > > But I guess there is a reason for sz_max being 0. Rob, Richard (commit > signer of this code) do you know why? Could you explain? > >> Anyway, this change still removes some bounds checking. Are they >> important? Do they need to be added back? >> > > The edge-slowdown and the vddmac values are compared against a const > array so we´re fine with those ones. > > For the led-X-mode, I added a constant for supported modes that gets > checked when retrieving the DT property. So we´re fine here too. > > Quentin >