Which bugs have you got with your OS?

[john765]

What kind of bugs did you get with your OS? Things repeating? Your OS reboots your machine every time the OS loads?
Share them here.

[iansjack]

The current bug I'm working on is a problem with filling in the relocation values when loading an executable that uses more than one dynamic library. It's got to be something really simple.

[zaval]

The current bug I'm working on is a problem with filling in the relocation values when loading an executable that uses more than one dynamic library. It's got to be something really simple.

Are you using PE for your images?




My current problem is figuring out what exactly Ingenic meant creating their manual. in the part on memory map, because it's awfully confusing. you need to guess what that arrow in the drawing does mean, to get the idea on proper memory map.
it's not a bug yet, because I didn't run SDRAM init code yet, even though it is written, but I suspect, this incostistency and incompleteness of the documentation will be really a great bug source.

[alexfru]

My current problem is figuring out what exactly Ingenic meant creating their manual. in the part on memory map, because it's awfully confusing. you need to guess what that arrow in the drawing does mean, to get the idea on proper memory map.
it's not a bug yet, because I didn't run SDRAM init code yet, even though it is written, but I suspect, this incostistency and incompleteness of the documentation will be really a great bug source.

The ingenic doc mentions that the TLB is the same as in MIPS32 4KE, meaning MIPS32 4KEc. 4KE should only differ from 4K in DSP instructions and R2 instructions. So, you should download the 4K docs from the Imagination website. As for the arrow, if I understood it right, it means that the same RAM is accessible in two ranges of the physical address space.

[iansjack]

I'm using elf64 images. I think I'm looking in the wrong .so file for some functions. It's just a matter is single-stepping my code to find exactly where I'm going wrong.

[eryjus]

What kind of bugs...?

Are you defining a "bug" as "trying to work through and complete some bit of code that you just wrote and would not call complete because it doesn't work as intended yet"? If so, I am parsing the MB2 information block and looking for framebuffer information.

Or, are you defining a "bug" as "something other than the current task that has thought to have worked but has now found to have some error that needs to be fixed"? If so, I do not have one that I know of yet (not because I write perfect code; this is simply because I have not laid enough code down to have one in this current incarnation or it still too early to have found it yet).

While I would refer to the first instance as "debugging" I do not consider them to be bugs in the code yet; rather, it is simply part of the development lifecycle. I don't even bother to capture them as bugs in my issue tracker. It is not until I have pushed the code to GitHub and mentally moved on from the topic as complete that anything new I find is considered a bug.

[zaval]

The ingenic doc mentions that the TLB is the same as in MIPS32 4KE, meaning MIPS32 4KEc. 4KE should only differ from 4K in DSP instructions and R2 instructions. So, you should download the 4K docs from the Imagination website. As for the arrow, if I understood it right, it means that the same RAM is accessible in two ranges of the physical address space.

It's so fun that you got it what arrow I was talking about instantly. Not that Ingenic's manuals are so popular slipslop.
Yes, I understood this the same. But there is a problem with this. According to the DDRC spec, one could assign available 1GB of SDRAM so that, it will be fitting into the range 0x00000000-0x40000000 (not inclusive). If you choose BASE and MASK fields 0x00 and 0x80 respectively, you kind of should get yet another alias range 0x40000000-0x80000000 pointing to the same locations, because bit 30 is don't care for DDRC. (setting the mask to 0xc0, would not let aliases, and you are going to loose 256 MB of space due to the reservation of 0x10000000-0x20000000 for devices). These aliases should be an easy way of accessing all SDRAM. But if that arrow works as we are thinking, it introduces a significant problem there. Because then ranges from 0x00000000-0x10000000 and 0x20000000-0x30000000 are aliases too. But they can not be aliases according to the chosen DDRC SDRAM mapping, which is valid! That was the "incostistency" I was talking about. In fact, with this possible folding it's not clear at all how it's going to work. how to address all the 1GB of SDRAM.

PS. Not to mention they brought yet another not mapped (magically mapped) undocumented region in the above part of kseg3 (0xf4000000-0xf5000000, where TCSM is mapped to) and forgot to elaborate on that.
Basically, after SDRAM init, the first thing needed is checking for possible overlapping between 0x00 and 0x20 ranges.

[zaval]

I'm using elf64 images. I think I'm looking in the wrong .so file for some functions. It's just a matter is single-stepping my code to find exactly where I'm going wrong.

I thought you have a base relocation issue, having multiple libraries linked at the same base, and thus needing to go through relocation process, that's why i asked about PE.
still i have little knowledge on elf with this respect, I thought its .so files use PIC. and that they have relocations only at link time.

[hgoel]

Been dealing with race conditions in my kernel, it's getting better but despite spending a lot of time on it, I can't get SMP to be reliable just yet. Fun stuff.