This change should really posted to OpenZFS too.

Thanks!

In D43914#1001826, @imp wrote:

In D43914#1001802, @tsoome wrote:

alloca or local variable does allocate from stack (which may be in collision with bss), the opther option there to get low memory allocation is to use bio_alloc() and bio_free().

Pnp nodes are like 32 bytes... and must be in the lower MB.

alloca or local variable does allocate from stack (which may be in collision with bss), the opther option there to get low memory allocation is to use bio_alloc() and bio_free().

The trick question there is, the bootonce property is supposed to have generic boot descriptor, it does not have to be just zfs dataset name.... of course, if you have pools, then most likely you only do use zfs datasets.

hm, do we actually need this preprocessor #if statement?

In D40011#911068, @emaste wrote:

Yeah. In the past I would have argued that we should work on support it (back when Intel had viable contemporary 32-bit hw and a 32-bit Tianocore to go with it) but at this point i386's future overall is not clear and we're not going to put effort into this.

In D39799#906388, @imp wrote:

so how was all this working before?
or is SEGDESC_t not all that important?

In D39323#895597, @manu wrote:

In D39323#895586, @hselasky wrote:

I think doing a functional function trace is mandatory before this commit, seeing what code is calling into what code, and what locks are in use.

If you need help ask, @zlei .

What @gbe reports clearly shows we can do better than just pushing changes and waiting for others to fix them.

--HPS

Nobody here said anything about "waiting for others to fix them". I hope that you understand that we cannot test every configuration possible.
And, sure, since now we have a report that it doesn't work on hyper-v nothing will be commited before the problem is understood and fixed.