netdev - Re: [PATCH net-next 3/5] r8152: add MCU typed read/write functions

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 for Android: free password hash cracker in your pocket

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

Message-ID: <20201105105418.555d6e54@kernel.org>
Date:   Thu, 5 Nov 2020 10:54:18 +0100
From:   Marek Behún <kabel@...nel.org>
To:     Vladimir Oltean <olteanv@...il.com>
Cc:     netdev@...r.kernel.org, linux-usb@...r.kernel.org,
        Hayes Wang <hayeswang@...ltek.com>
Subject: Re: [PATCH net-next 3/5] r8152: add MCU typed read/write functions

On Wed, 4 Nov 2020 14:14:24 +0200
Vladimir Oltean <olteanv@...il.com> wrote:

> On Wed, Nov 04, 2020 at 12:10:53PM +0100, Marek Behún wrote:
> > > I'm not sure it's worth the change :(
> > > Let's put it another way, your diffstat has 338 insertions and 335
> > > deletions. Aka you're saving 3 lines overall.
> > > With this new approach that doesn't use token concatenation at all,
> > > you're probably not saving anything at all.
> > > Also, I'm not sure that you need to make the functions inline. The
> > > compiler should be smart enough to not generate functions for
> > > usb_ocp_read_byte etc. You can check with
> > > "make drivers/net/usb/r8152.lst".  
> > 
> > Vladimir, the purpose of this patch isn't to save lines, but to save us
> > from always writing MCU_TYPE_USB / MCU_TYPE_PLA.
> > It just transforms forms of
> >   ocp_read_word(tp, MCU_TYPE_USB, idx);
> >   ocp_write_dword(tp, MCU_TYPE_PLA, idx, val);
> > into
> >   usb_ocp_read_word(tp, idx);
> >   pla_ocp_write_dword(tp, idx, val);
> > 
> > The fifth patch of this series saves lines by adding _modify functions,
> > to transform
> >   val = *_read(idx);
> >   val &= ~clr;
> >   val |= set;
> >   *_write(idx, val);
> > into
> >   *_modify(idx, clr, set);
> >   
> 
> So if the point isn't to save lines, then why don't you go for something
> trivial?
> 
> static void ocp_modify_byte(struct r8152 *tp, u16 type, u16 index, u8 clr,
> 			    u8 set)
> {
> 	u8 val = ocp_read_byte(tp, type, index);
> 
> 	ocp_write_byte(tp, type, index, (val & ~clr) | set);
> }
> 
> static void ocp_modify_word(struct r8152 *tp, u16 type, u16 index, u16 clr,
> 			    u16 set)
> {
> 	u16 val = ocp_read_word(tp, type, index);
> 
> 	ocp_write_word(tp, type, index, (val & ~clr) | set);
> }
> 
> static void ocp_modify_dword(struct r8152 *tp, u16 type, u16 index, u32 clr,
> 			     u32 set)
> {
> 	u32 val = ocp_read_dword(tp, type, index);
> 
> 	ocp_write_dword(tp, type, index, (val & ~clr) | set);
> }
> 
> #define pla_ocp_read_byte(tp, index)				\
> 	ocp_read_byte(tp, MCU_TYPE_PLA, index)
> #define pla_ocp_write_byte(tp, index, data)			\
> 	ocp_write_byte(tp, MCU_TYPE_PLA, index, data)
> #define pla_ocp_modify_byte(tp, index, clr, set)		\
> 	ocp_modify_byte(tp, MCU_TYPE_PLA, index, clr, set)
> #define pla_ocp_read_word(tp, index)				\
> 	ocp_read_word(tp, MCU_TYPE_PLA, index)
> #define pla_ocp_write_word(tp, index, data)			\
> 	ocp_write_word(tp, MCU_TYPE_PLA, index, data)
> #define pla_ocp_modify_word(tp, index, clr, set)		\
> 	ocp_modify_word(tp, MCU_TYPE_PLA, index, clr, set)
> #define pla_ocp_read_dword(tp, index)				\
> 	ocp_read_dword(tp, MCU_TYPE_PLA, index)
> #define pla_ocp_write_dword(tp, index, data)			\
> 	ocp_write_dword(tp, MCU_TYPE_PLA, index, data)
> #define pla_ocp_modify_dword(tp, index, clr, set)		\
> 	ocp_modify_dword(tp, MCU_TYPE_PLA, index, clr, set)
> 
> #define usb_ocp_read_byte(tp, index)				\
> 	ocp_read_byte(tp, MCU_TYPE_USB, index)
> #define usb_ocp_write_byte(tp, index, data)			\
> 	ocp_write_byte(tp, MCU_TYPE_USB, index, data)
> #define usb_ocp_modify_byte(tp, index, clr, set)		\
> 	ocp_modify_byte(tp, MCU_TYPE_USB, index, clr, set)
> #define usb_ocp_read_word(tp, index)				\
> 	ocp_read_word(tp, MCU_TYPE_USB, index)
> #define usb_ocp_write_word(tp, index, data)			\
> 	ocp_write_word(tp, MCU_TYPE_USB, index, data)
> #define usb_ocp_modify_word(tp, index, clr, set)		\
> 	ocp_modify_word(tp, MCU_TYPE_USB, index, clr, set)
> #define usb_ocp_read_dword(tp, index)				\
> 	ocp_read_dword(tp, MCU_TYPE_USB, index)
> #define usb_ocp_write_dword(tp, index, data)			\
> 	ocp_write_dword(tp, MCU_TYPE_USB, index, data)
> #define usb_ocp_modify_dword(tp, index, clr, set)		\
> 	ocp_modify_dword(tp, MCU_TYPE_USB, index, clr, set)
> 
> To my eyes this is easier to digest.
> 
> That is, unless you want to go for function pointers and have separate
> structures for PLA and USB...

I thought that static inline functions are preferred to macros, since
compiler warns better if they are used incorrectly...