[blibble]

if it's unordered (like you'd expect a list of sockets to be) you can get O(1) removal for an array by swapping the end item into the removed slot

[dozzie]

...unless something else holds a pointer to the handle you're swapping with.

[staticassertion]

This is already a problem if you're removing an item, regardless of whether it's done with swap remove or not.

[toast0]

For the item you're removing, you are presumably removing it because all handles have been closed. But the item you're potentially swapping with may certainly have open handles.

[staticassertion]

True.

[blibble]

kernels don't tend to give userspace raw pointers into kernel memory

[dozzie]

But kernels tend to give raw pointers to other kernel code.

[blibble]

a good chunk of the fun in kernel code is maintaining the various layers of self-referential lists and dealing with the cleanup