Re: [broonie-ci:20220430_marex_regmap_add_bulk_read_write_callbacks_into_regmap_config 1/1] drivers/base/regmap/regmap.c:1842 _regmap_raw_write_impl() error: we previously assumed 'map->bus' could be null (see line 1770)

[Marek Vasut]

On 5/6/22 09:12, Dan Carpenter wrote:
> tree:   https://git.kernel.org/pub/scm/linux/kernel/git/broonie/ci.git 20220430_marex_regmap_add_bulk_read_write_callbacks_into_regmap_config
> head:   d77e745613680c54708470402e2b623dcd769681
> commit: d77e745613680c54708470402e2b623dcd769681 [1/1] regmap: Add bulk read/write callbacks into regmap_config
> config: m68k-randconfig-m031-20220505 (https://download.01.org/0day-ci/archive/20220506/202205060518.hir67qzc-lkp@xxxxxxxxx/config)
> compiler: m68k-linux-gcc (GCC) 11.3.0
>
> If you fix the issue, kindly add following tag as appropriate
> Reported-by: kernel test robot <lkp@xxxxxxxxx>
> Reported-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
>
> New smatch warnings:
> drivers/base/regmap/regmap.c:1842 _regmap_raw_write_impl() error: we previously assumed 'map->bus' could be null (see line 1770)
>
> vim +1842 drivers/base/regmap/regmap.c

[...]

> 0074f3f2b1e43d Colin Foster        2022-03-13  1754  	reg += map->reg_base;
> 86fc59ef818beb Colin Foster        2022-03-13  1755  	reg >>= map->format.reg_downshift;
> d939fb9a78b474 Marc Reilly         2012-03-16  1756  	map->format.format_reg(map->work_buf, reg, map->reg_shift);
> f50e38c9966076 Tony Lindgren       2016-09-15  1757  	regmap_set_work_buf_flag_mask(map, map->format.reg_bytes,
> f50e38c9966076 Tony Lindgren       2016-09-15  1758  				      map->write_flag_mask);
> 6f306441e97f8f Lars-Peter Clausen  2011-09-05  1759
> 651e013e3ce6c0 Mark Brown          2013-10-08  1760  	/*
> 651e013e3ce6c0 Mark Brown          2013-10-08  1761  	 * Essentially all I/O mechanisms will be faster with a single
> 651e013e3ce6c0 Mark Brown          2013-10-08  1762  	 * buffer to write.  Since register syncs often generate raw
> 651e013e3ce6c0 Mark Brown          2013-10-08  1763  	 * writes of single registers optimise that case.
> 651e013e3ce6c0 Mark Brown          2013-10-08  1764  	 */
> 651e013e3ce6c0 Mark Brown          2013-10-08  1765  	if (val != work_val && val_len == map->format.val_bytes) {
> 651e013e3ce6c0 Mark Brown          2013-10-08  1766  		memcpy(work_val, val, map->format.val_bytes);
> 651e013e3ce6c0 Mark Brown          2013-10-08  1767  		val = work_val;
> 651e013e3ce6c0 Mark Brown          2013-10-08  1768  	}
> 651e013e3ce6c0 Mark Brown          2013-10-08  1769
> d77e745613680c Marek Vasut         2022-04-30 @1770  	if (map->async && map->bus && map->bus->async_write) {
>                                                                            ^^^^^^^^
> Do we really need to check map->bus?

Here I think so, because bus can be NULL.

[...]

> c6b570d97c0e77 Philipp Zabel       2015-03-09  1831  	trace_regmap_hw_write_start(map, reg, val_len / map->format.val_bytes);
> fb2736bbaee0e7 Mark Brown          2011-07-24  1832
> 2547e201b3693f Mark Brown          2011-07-20  1833  	/* If we're doing a single register write we can probably just
> 2547e201b3693f Mark Brown          2011-07-20  1834  	 * send the work_buf directly, otherwise try to do a gather
> 2547e201b3693f Mark Brown          2011-07-20  1835  	 * write.
> 2547e201b3693f Mark Brown          2011-07-20  1836  	 */
> 0d509f2b112b21 Mark Brown          2013-01-27  1837  	if (val == work_val)
> d77e745613680c Marek Vasut         2022-04-30  1838  		ret = map->write(map->bus_context, map->work_buf,
> 82159ba8e6ef8c Mark Brown          2012-01-18  1839  				 map->format.reg_bytes +
> 82159ba8e6ef8c Mark Brown          2012-01-18  1840  				 map->format.pad_bytes +
> 82159ba8e6ef8c Mark Brown          2012-01-18  1841  				 val_len);
> 2547e201b3693f Mark Brown          2011-07-20 @1842  	else if (map->bus->gather_write)
>                                                                   ^^^^^^^^
> It's not checked here

This is a bug indeed.