copyout page fault

Barney

Member

Reaction score: 1
Messages: 59

In 12.1B I'm getting a page fault panic on a large copyout; something I never saw with previous versions. It appears something has changed. Isnt the point of copyout to lock down the pages during the copy? It seems in 12 the page in user space isn't locked down. Anyone know what's changed?
 

yuripv

Well-Known Member

Reaction score: 112
Messages: 262

More details are definitely welcome. May be you could provide a code snippet showing what exactly you are doing?
 
OP
OP
B

Barney

Member

Reaction score: 1
Messages: 59

I've found some other threads where it's outlined that stuff that's worked for 20 years no longer works because none of the corporate sponsors care about it. One of the "improvements" in 12.
 

shkhln

Aspiring Daemon

Reaction score: 278
Messages: 858

Hey, we are just trying to extract any useful information from you before you are getting banned. Go with the flow, dammit.
 

ralphbsz

Daemon

Reaction score: 1,250
Messages: 1,999

In 12.1B ...
So you are running an unknown version (I have no idea what "B" means), and furthermore an as yet unreleased version. That means that anything that breaks is to be expected.

I'm getting a page fault panic on ...
If you want assistance debugging it, we need WAY MORE information. At the minimum a stack trace with register dumps, and a description of what the system was doing. If you are running a pre-release version, debugging it probably also requires showing us the source code of where the problem occurred.

Isnt the point of copyout to lock down the pages during the copy?
Describing the purpose of a significantly large part of the kernel in a half dozen words is not going to work. It's way more complex that that.

It seems in 12 the page in user space isn't locked down.
If in general the pages weren't pinned or locked during the page out IO operation, then fundamentally nothing would work. That seems exceedingly unlikely. It is also the kind of elementary mistake that no seasoned kernel developer would make, and that would not get through reviews. Now, it is possible that there is a bug in the logic (wouldn't be the first one, and won't be the last one), but such a generalization makes it laughable.

I've found some other threads where it's outlined that stuff that's worked for 20 years no longer works because none of the corporate sponsors care about it. One of the "improvements" in 12.
Sorry, if you want to make paranoid accusations, you need to give us some evidence. Extraordinary claims require extraordinary evidence. The idea that no corporate sponsors care about basic functionality of the OS (which paging is) is nonsense. As is your mental concept that FreeBSD development is driven by decisions of corporate sponsors. I don't know whether it has become clear to you that this is not Linux, and the sociology and economics of Linux does not apply here.

Please show us which threads, what has been broken, and which "corporate sponsors" used to keep it running. Failing that, my explanation for your behavior here in this thread is that you are just being disruptive and trolling.
 

shkhln

Aspiring Daemon

Reaction score: 278
Messages: 858

That might be it, however OP's description of their problem (copyout "misbehaving") is inconsistent with the problem in that thread, namely access to userspace memory without copyin/copyout.
 
OP
OP
B

Barney

Member

Reaction score: 1
Messages: 59

So you are running an unknown version (I have no idea what "B" means), and furthermore an as yet unreleased version. That means that anything that breaks is to be expected.
B = Beta. Sorry if this threw you.

It doesn't work anymore. So I'm doing it differently. Nothing to debug or discuss further. I'm not sure why you're getting all angry. The problem with forums is that by the time someone responds, you've already either figured it out or moved in another direction.

The idea that so many things are broken in a Beta2 that it's not worth discussing things that don't work is utterly ridiculous. That's what Betas are for: to find remaining things that are broken.
 
OP
OP
B

Barney

Member

Reaction score: 1
Messages: 59

That might be it, however OP's description of their problem (copyout "misbehaving") is inconsistent with the problem in that thread, namely access to userspace memory without copyin/copyout.
Perhaps the OP did some research and testing on SMAP outside of the forum?
 

shkhln

Aspiring Daemon

Reaction score: 278
Messages: 858

Perhaps OP should learn how to debug software by isolating malfunctioning parts, rather than relying on wishful thinking.
 
OP
OP
B

Barney

Member

Reaction score: 1
Messages: 59

Perhaps OP should learn how to debug software by isolating malfunctioning parts, rather than relying on wishful thinking.
perhaps the OP doesn't need or want your advice or help? I'll assume your English is lacking because I have no idea what you're talking about. I debugged and solved the problem without your help. Sorry if that upsets you.
 
OP
OP
B

Barney

Member

Reaction score: 1
Messages: 59

Demanding that people continue to provide details for something they no longer care about is mental illness defined.
 
Top