vfio/pci: Lazy PBA emulation

The PCI spec recommends devices use additional alignment for MSI-X
data structures to allow software to map them to separate processor
pages.  One advantage of doing this is that we can emulate those data
structures without a significant performance impact to the operation
of the device.  Some devices fail to implement that suggestion and
assigned device performance suffers.

One such case of this is a Mellanox MT27500 series, ConnectX-3 VF,
where the MSI-X vector table and PBA are aligned on separate 4K
pages.  If PBA emulation is enabled, performance suffers.  It's not
clear how much value we get from PBA emulation, but the solution here
is to only lazily enable the emulated PBA when a masked MSI-X vector
fires.  We then attempt to more aggresively disable the PBA memory
region any time a vector is unmasked.  The expectation is then that
a typical VM will run entirely with PBA emulation disabled, and only
when used is that emulation re-enabled.

Reported-by: Shyam Kaushik <shyam.kaushik@gmail.com>
Tested-by: Shyam Kaushik <shyam.kaushik@gmail.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
3 files changed
tree: 357b84f217fbc385239e12863fe87eb514ceef2d
  1. audio/
  2. backends/
  3. block/
  4. bsd-user/
  5. contrib/
  6. crypto/
  7. default-configs/
  8. disas/
  9. docs/
  10. fpu/
  11. fsdev/
  12. gdb-xml/
  13. hw/
  14. include/
  15. io/
  16. libdecnumber/
  17. linux-headers/
  18. linux-user/
  19. migration/
  20. nbd/
  21. net/
  22. pc-bios/
  23. po/
  24. qapi/
  25. qga/
  26. qobject/
  27. qom/
  28. replay/
  29. roms/
  30. scripts/
  31. slirp/
  32. stubs/
  33. target-alpha/
  34. target-arm/
  35. target-cris/
  36. target-i386/
  37. target-lm32/
  38. target-m68k/
  39. target-microblaze/
  40. target-mips/
  41. target-moxie/
  42. target-openrisc/
  43. target-ppc/
  44. target-s390x/
  45. target-sh4/
  46. target-sparc/
  47. target-tilegx/
  48. target-tricore/
  49. target-unicore32/
  50. target-xtensa/
  51. tcg/
  52. tests/
  53. trace/
  54. ui/
  55. util/
  56. .dir-locals.el
  57. .exrc
  58. .gitignore
  59. .gitmodules
  60. .mailmap
  61. .travis.yml
  62. accel.c
  63. aio-posix.c
  64. aio-win32.c
  65. arch_init.c
  66. async.c
  67. balloon.c
  68. block.c
  69. blockdev-nbd.c
  70. blockdev.c
  71. blockjob.c
  72. bootdevice.c
  73. bt-host.c
  74. bt-vhci.c
  75. Changelog
  76. CODING_STYLE
  77. configure
  78. COPYING
  79. COPYING.LIB
  80. cpu-exec-common.c
  81. cpu-exec.c
  82. cpus.c
  83. cputlb.c
  84. device-hotplug.c
  85. device_tree.c
  86. disas.c
  87. dma-helpers.c
  88. dump.c
  89. exec.c
  90. gdbstub.c
  91. HACKING
  92. hmp-commands-info.hx
  93. hmp-commands.hx
  94. hmp.c
  95. hmp.h
  96. iohandler.c
  97. ioport.c
  98. iothread.c
  99. kvm-all.c
  100. kvm-stub.c
  101. LICENSE
  102. main-loop.c
  103. MAINTAINERS
  104. Makefile
  105. Makefile.objs
  106. Makefile.target
  107. memory.c
  108. memory_mapping.c
  109. module-common.c
  110. monitor.c
  111. numa.c
  112. os-posix.c
  113. os-win32.c
  114. page_cache.c
  115. qapi-schema.json
  116. qdev-monitor.c
  117. qdict-test-data.txt
  118. qemu-bridge-helper.c
  119. qemu-char.c
  120. qemu-doc.texi
  121. qemu-ga.texi
  122. qemu-img-cmds.hx
  123. qemu-img.c
  124. qemu-img.texi
  125. qemu-io-cmds.c
  126. qemu-io.c
  127. qemu-log.c
  128. qemu-nbd.c
  129. qemu-nbd.texi
  130. qemu-options-wrapper.h
  131. qemu-options.h
  132. qemu-options.hx
  133. qemu-seccomp.c
  134. qemu-tech.texi
  135. qemu-timer.c
  136. qemu.nsi
  137. qemu.sasl
  138. qjson.c
  139. qmp-commands.hx
  140. qmp.c
  141. qtest.c
  142. README
  143. rules.mak
  144. softmmu_template.h
  145. spice-qemu-char.c
  146. tcg-runtime.c
  147. tci.c
  148. thread-pool.c
  149. thunk.c
  150. tpm.c
  151. trace-events
  152. translate-all.c
  153. translate-all.h
  154. translate-common.c
  155. user-exec.c
  156. VERSION
  157. version.rc
  158. vl.c
  159. xen-common-stub.c
  160. xen-common.c
  161. xen-hvm-stub.c
  162. xen-hvm.c
  163. xen-mapcache.c