Approved.

Looks ok to me with the man page nits fixed.

Do you have some local modifications to test the ADDR_IS_KERNEL(va) code path in pmap_enter_l2()?

Don't count unswappable pages as dirty. I believe this will address kib's concern
by side-stepping it: if a dirty page can't be swapped out, we simply don't count it
against the scan limit, so we will process as many pages as necessary.

• Rebase.
• Drop the change to permit clustering with dirty pages in the inactive queue. It's logically a separate change, and it was incomplete: on its own it may cause the swap pager to call vm_page_deactivate_noreuse() on pages that hadn't yet made a full trip through the inactive queue. Compare with vm_pageout_flush(), which only moves pages to the head of the inactive queue if they are in the laundry.

Thank you. I guess the problem arises when the directory is a mount point, and we need to upgrade the lock?

Modify the arm64 pmap to avoid clearing ATTR_AF on wired PTEs as well.

In D50820#1163168, @aokblast wrote:

Please use this patch to merge. https://github.com/aokblast/freebsd-src/commit/ff36a394a7e84e2576db83f1ef5f28cd61e36366.patch Thanks!

BTW there are some test failures when this change is applied. e.g.,

I verified that the test failures are gone now, thank you.

Approved, but see the inline comment.

Perhaps modify extattr.2 and extattr.3 to reference this page?

In D51188#1168819, @zldrobit_gmail.com wrote:

@markj I found that we should also delete st_value = 0 in dt_module_symsort32/64(), otherwise dmp->dm_aslen would always be lesser than dmp->dm_asrsv by at least one. As a result, dt_module_symsort32/64() ignore the last element in dmp->dm_asmap and dt_module_symaddr32/64() wouldn't find it. I had to update the patch accordingly. Sorry for interrupting the test suite procedure.

In D51188#1168819, @zldrobit_gmail.com wrote:

@markj I found that we should also delete st_value = 0 in dt_module_symsort32/64(), otherwise dmp->dm_aslen would always be lesser than dmp->dm_asrsv by at least one. As a result, dt_module_symsort32/64() ignore the last element in dmp->dm_asmap and dt_module_symaddr32/64() wouldn't find it. I had to update the patch accordingly. Sorry for interrupting the test suite procedure.

I'll commit this once I've gone through the test suite.

Apply feedback from Conrad.

Bump __FreeBSD_version.

This change looks reasonable to me. I think the problem with st_value = 0 only applies to ET_REL KLDs.

Please be sure to include a Fixes: tag in the commit message.

Looks pretty good to me.

This is looking pretty good, I just had one question about compatibility, the other comments are just nits.

In D50740#1167971, @aokblast wrote:

In D50740#1167967, @markj wrote:

In libusb20_desc.h I see a macro LIBUSB20_BOS_DESCRIPTOR which references the old field name. I'm not really sure what that macro is doing--should that be updated as well?

The BOS_DESCRIPTOR in libusb20 is a dedicated structure for decode in libusb20 and the definition resides inside the libusb20_desc.c. Thus there is no reference from libusb10 structure.

For libusb10, we require a xfer and copy it into buffer directly. It will not be parsed by libusb20.

In libusb20_desc.h I see a macro LIBUSB20_BOS_DESCRIPTOR which references the old field name. I'm not really sure what that macro is doing--should that be updated as well?