Was this a stupid mistake?

[pcmattman]

OK, I was coding my fopen function to fix some bugs, and all of a sudden my kfree tells me it's trying to free non-heap memory. I think to myself, 'why?'...

Here's why:

Code: Select all

char* fbuff = (char*) kmalloc( 256 );
fbuff = (char*) PadFilename( (uchar_t*) filename );

10 minutes I sat there until I thought of saving the first address...

Code: Select all

char* fbuff = (char*) kmalloc( 256 );
uint_t fbuff_addr = (uint_t) fbuff; /** we trash the fbuff pointer, and I don't yet know how **/
fbuff = (char*) PadFilename( (uchar_t*) filename );

And then I figured it out:

Code: Select all

char* fbuff = (char*) kmalloc( 256 );
uint_t fbuff_addr = (uint_t) fbuff; /** we trash the fbuff pointer, and I don't yet know how **/
fbuff = (char*) PadFilename( (uchar_t*) filename ); /** just figured it out... **/

I just overwrote the pointer in the next line of code

So, was this a stupid mistake?

[Zacariaz]

i have no clue what you talking about, but isnt any mistake kinda stupid?

[distantvoices]

Stuff of that sort always happens.

It's just mistakes, nothing bad.

Learn from this experience, young padawan, and don't hesitate to continue. You'll gonna make more mistakes. They'll grow in level, of course, and hide way better.

Stay safe

[pcmattman]

One bug I had once had each memory block allocated by my malloc running all over the place, so data was in the wrong place - corrupted memory.

Now that's all fixed

[distantvoices]

Reminds me of some hidden and really hard to find bug inside my paging stuff: I 've just allocated pages for pagetables and assigned them. Of course, they weren't zeroed out, so the mmu actually FOUND some memory to translate a given address the cpu has put on the address bus.

unfortunately, this caused the oddest of bugs to show up and none of them were reproduceable - initially. After a while of watching I 've been able to deduce a certain pattern, and from that I 've drawn the conlusion, that my memory management 's been buggy.

Well, and once that 's been clear, smashing the f***ing bug 's been a walk in the park.

The difficult stuff is discovering where the bloody heck the bug is hidden.

[Aali]

the best bugs come from not flushing the TLB properly, though

everything looks okay, every pagetable is correct, and you still trash memory pseudo-randomly

[inflater]

Everybody are making mistakes, humans, computers and maybe animals too.

inflater

[Candy]

Best thing you can do when making a mistake is to note what went wrong, how you figured that out and how you fixed it. I've been able to use a number of things from here - nonobvious non-normal things - in my work. The two most odd things that both my OS and my work had was one triple fault problem (spurious interrupt handler not installed, for both) and one executable-not-fully-loaded problem (which was pretty much undebuggable, in either case, except that I had a sharp lookout for nice numbers and a lot of guts saying "it's this". OS hit the limit at 32k, work item hit at 1M).

[os64dev]

Everybody are making mistakes, humans, computers and maybe animals too.
inflater

Computers don't make mistakes, only the people using them.

[AJ]

What about changes in bit patterns due to cosmic rays? (I know, I know...happens all the time, doesn't it...)

[Brendan]

Hi,

What about changes in bit patterns due to cosmic rays? (I know, I know...happens all the time, doesn't it...)

That would be caused by inadaquate shielding (a mistake made by human designers)...


Cheers,

Brendan