6.18-stable review patch. If anyone has any objections, please let me know.
------------------
From: Charles Mirabile cmirabil@redhat.com
[ Upstream commit 23b2d2fb136959fd0a8e309c70be83d9b8841c7e ]
When booting with KASAN enabled the following splat is encountered during probe of the k1 clock driver:
UBSAN: array-index-out-of-bounds in drivers/clk/spacemit/ccu-k1.c:1044:16 index 0 is out of range for type 'clk_hw *[*]' CPU: 0 UID: 0 PID: 1 Comm: swapper/0 Not tainted 6.18.0-rc5+ #1 PREEMPT(lazy) Hardware name: Unknown Unknown Product/Unknown Product, BIOS 2022.10spacemit 10/01/2022 Call Trace: [<ffffffff8002b628>] dump_backtrace+0x28/0x38 [<ffffffff800027d2>] show_stack+0x3a/0x50 [<ffffffff800220c2>] dump_stack_lvl+0x5a/0x80 [<ffffffff80022100>] dump_stack+0x18/0x20 [<ffffffff800164b8>] ubsan_epilogue+0x10/0x48 [<ffffffff8099034e>] __ubsan_handle_out_of_bounds+0xa6/0xa8 [<ffffffff80acbfa6>] k1_ccu_probe+0x37e/0x420 [<ffffffff80b79e6e>] platform_probe+0x56/0x98 [<ffffffff80b76a7e>] really_probe+0x9e/0x350 [<ffffffff80b76db0>] __driver_probe_device+0x80/0x138 [<ffffffff80b76f52>] driver_probe_device+0x3a/0xd0 [<ffffffff80b771c4>] __driver_attach+0xac/0x1b8 [<ffffffff80b742fc>] bus_for_each_dev+0x6c/0xc8 [<ffffffff80b76296>] driver_attach+0x26/0x38 [<ffffffff80b759ae>] bus_add_driver+0x13e/0x268 [<ffffffff80b7836a>] driver_register+0x52/0x100 [<ffffffff80b79a78>] __platform_driver_register+0x28/0x38 [<ffffffff814585da>] k1_ccu_driver_init+0x22/0x38 [<ffffffff80023a8a>] do_one_initcall+0x62/0x2a0 [<ffffffff81401c60>] do_initcalls+0x170/0x1a8 [<ffffffff81401e7a>] kernel_init_freeable+0x16a/0x1e0 [<ffffffff811f7534>] kernel_init+0x2c/0x180 [<ffffffff80025f56>] ret_from_fork_kernel+0x16/0x1d8 [<ffffffff81205336>] ret_from_fork_kernel_asm+0x16/0x18 ---[ end trace ]---
This is bogus and is simply a result of KASAN consulting the `.num` member of the struct for bounds information (as it should due to `__counted_by`) and finding 0 set by kzalloc() because it has not been initialized before the loop that fills in the array. The easy fix is to just move the line that sets `num` to before the loop that fills the array so that KASAN has the information it needs to accurately conclude that the access is valid.
Fixes: 1b72c59db0add ("clk: spacemit: Add clock support for SpacemiT K1 SoC") Tested-by: Yanko Kaneti yaneti@declera.com Signed-off-by: Charles Mirabile cmirabil@redhat.com Reviewed-by: Alex Elder elder@riscstar.com Reviewed-by: Troy Mitchell troy.mitchell@linux.spacemit.com Reviewed-by: Yixun Lan dlan@gentoo.org Signed-off-by: Stephen Boyd sboyd@kernel.org Signed-off-by: Sasha Levin sashal@kernel.org --- drivers/clk/spacemit/ccu-k1.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/clk/spacemit/ccu-k1.c b/drivers/clk/spacemit/ccu-k1.c index f5a9fe6ba1859..4761bc1e3b6e6 100644 --- a/drivers/clk/spacemit/ccu-k1.c +++ b/drivers/clk/spacemit/ccu-k1.c @@ -1018,6 +1018,8 @@ static int spacemit_ccu_register(struct device *dev, if (!clk_data) return -ENOMEM;
+ clk_data->num = data->num; + for (i = 0; i < data->num; i++) { struct clk_hw *hw = data->hws[i]; struct ccu_common *common; @@ -1044,8 +1046,6 @@ static int spacemit_ccu_register(struct device *dev, clk_data->hws[i] = hw; }
- clk_data->num = data->num; - ret = devm_of_clk_add_hw_provider(dev, of_clk_hw_onecell_get, clk_data); if (ret) dev_err(dev, "failed to add clock hardware provider (%d)\n", ret);