Hi Tom, On Fri, 21 Nov 2025 at 12:27, Tom Rini <[email protected]> wrote: > > On Fri, Nov 21, 2025 at 12:09:31PM -0700, Simon Glass wrote: > > Hi Marek, > > > > On Fri, 21 Nov 2025 at 10:53, Marek Vasut <[email protected]> wrote: > > > > > > On 11/21/25 6:35 PM, Simon Glass wrote: > > > > > > Hello Simon, > > > > > > >> To be made to clk_fixed_rate_ofdata_to_plat_(...) in > > > >> drivers/clk/clk_fixed_rate.c as it does not currently call > > > >> dev_has_ofnode(dev). > > > > > > > > Ah OK. So the call stack is something like: > > > > > > > > dev_read_u32_default() > > > > dev_ofnode() - returns ofnode_null > > > > ofnode_read_u32_default() > > > > ofnode_read_u32_index > > > > fdt_getprop(NULL, -1, ...) // guess > > > > > > > > And I see that I put assert() in some of the ofnode_read..() functions > > > > because I don't have tests for passing an invalid ofnode. > > > > > > > > So perhaps fdt_offset_ptr_() should check for NULL and return? I > > > > haven't traced it though. > > > > > > In previous patch > > > > > > [PATCH 3/3] libfdt: Check fdt_offset_ptr() return value unconditionally > > > > > > that approach seems to have been rejected, and these dev_read_*() fixes > > > were implemented as a replacement. > > > > I don't think it was rejected...but I think the commit message should > > better explain the problem, e.g. using some of the info above. > > > > The only choice is whether to add the check unconditionally (which I > > think is fine for now) or to add a new type of assumption. > > Your feedback here does not seem to match up with your feedback in the > patch in question: > https://lore.kernel.org/all/caflsztg6df1dmjjdh8-2zjuf6i9+zofvwnsxnrs_pbq3cdn...@mail.gmail.com/#t
Well in my defence the cover letter and that patch didn't give me much of a clue. No mention of OF_PLATDATA or that there is no actual FDT. I thought Marek was saying that libfdt no longer accepted negative offsets! Regards, Simon
