On Tue, Sep 08, 2020 at 09:53:11PM +0200, Pavel Machek wrote:
> Hi!
> 
> > On machines with much memory (> 2 TByte) and log_mtts_per_seg == 0, a
> > max_order of 31 will be passed to mlx_buddy_init(), which results in
> > s = BITS_TO_LONGS(1 << 31) becoming a negative value, leading to
> > kvmalloc_array() failure when it is converted to size_t.
> > 
> >   mlx4_core 0000:b1:00.0: Failed to initialize memory region table, aborting
> >   mlx4_core: probe of 0000:b1:00.0 failed with error -12
> > 
> > Fix this issue by changing the left shifting operand from a signed literal 
> > to
> > an unsigned one.
> 
> Will we still have problems with > 4 TByte machines?

AFAIK we're safe since max_buddy is calculated as such

        /* In drivers/net/ethernet/mellanox/mlx4/mr.c */
        err = mlx4_buddy_init(&mr_table->mtt_buddy,
                              ilog2((u32)dev->caps.num_mtts /
                              (1 << log_mtts_per_seg)));

Also, num_mtts is capped at 2^31

        /* In drivers/net/ethernet/mellanox/mlx4/profile.c */
        /*
         * We want to scale the number of MTTs with the size of the
         * system memory, since it makes sense to register a lot of
         * memory on a system with a lot of memory.  As a heuristic,
         * make sure we have enough MTTs to cover twice the system
         * memory (with PAGE_SIZE entries).
         *
         * This number has to be a power of two and fit into 32 bits
         * due to device limitations, so cap this at 2^31 as well.
         * That limits us to 8TB of memory registration per HCA with
         * 4KB pages, which is probably OK for the next few months.
         */
        si_meminfo(&si);
        request->num_mtt =
                roundup_pow_of_two(max_t(unsigned, request->num_mtt,
                                         min(1UL << (31 - log_mtts_per_seg),
                                             (si.totalram << 1) >> 
log_mtts_per_seg)));

Best,
Shung-Hsi Yu

> Should the computation be done in u64?
>
> Best regards,
>                                                                       Pavel
> 
> > Fixes: 225c7b1feef1 ("IB/mlx4: Add a driver Mellanox ConnectX InfiniBand 
> > adapters")
> > Signed-off-by: Shung-Hsi Yu <shung-hsi...@suse.com>
> > Signed-off-by: David S. Miller <da...@davemloft.net>
> > Signed-off-by: Sasha Levin <sas...@kernel.org>
> 
> > +++ b/drivers/net/ethernet/mellanox/mlx4/mr.c
> > @@ -114,7 +114,7 @@ static int mlx4_buddy_init(struct mlx4_buddy *buddy, 
> > int max_order)
> >             goto err_out;
> >  
> >     for (i = 0; i <= buddy->max_order; ++i) {
> > -           s = BITS_TO_LONGS(1 << (buddy->max_order - i));
> > +           s = BITS_TO_LONGS(1UL << (buddy->max_order - i));
> >             buddy->bits[i] = kvmalloc_array(s, sizeof(long), GFP_KERNEL | 
> > __GFP_ZERO);
> >             if (!buddy->bits[i])
> >                     goto err_out_free;
> 
> -- 
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures) 
> http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html


Reply via email to