Details
- Reviewers
markj - Commits
- rS352607: i386: implement sysctl vm.pmap.kernel_maps.
Diff Detail
- Repository
- rS FreeBSD src repository - subversion
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
| sys/i386/i386/pmap.c | ||
|---|---|---|
| 6197 ↗ | (On Diff #62396) | Is there a reason not to define and set pg_nx = 0 in non-PAE-capable kernels? |
| 6302 ↗ | (On Diff #62396) | The rounddown2 is not needed here. I used it in the amd64 variant since iteration may be restarted while examining the large map. Upon the restart, we might encounter a new large page, and sva will be misaligned. In this i386 variant there are no restarts, so this situation cannot occur. |
| sys/i386/i386/pmap.c | ||
|---|---|---|
| 6197 ↗ | (On Diff #62396) | It is a very minor optimization for non-PAE pmap. Actually, not. pg_nx has to be global, and then its type is not compatible between PAE and non-PAE pmaps, so it should not be touched by non-PAE at all. Note that PAE and non-PAE pmaps co-exist in the same kernel image. |
| sys/i386/i386/pmap.c | ||
|---|---|---|
| 6197 ↗ | (On Diff #62396) | Hmm, could we perhaps #define pg_nx 0u in pmap_nopae.h? |
| 6285 ↗ | (On Diff #62396) | In the amd64 sysctl there is an extra newline before each header, not just the first. I slightly prefer that style, but in any case we should try to be consistent. If you prefer to avoid the extra newlines, we should change amd64 too. |
| sys/i386/i386/pmap.c | ||
|---|---|---|
| 6197 ↗ | (On Diff #62396) | Perhaps. But this should be a separate change with the global pass over the pmap.c. |
| 6285 ↗ | (On Diff #62396) | The reason why I removed these newlines is because for i386 start of next herald does not delineate end of the previous run. I would require more work to properly mark the ends, and also I want to mark things like clean/exec/pipe submaps etc. This should be done for amd64 and i386 later. |