Browse Source

*** empty log message ***

master
Marc Alexander Lehmann 1 year ago
parent
commit
d6d5915bf9
2 changed files with 6 additions and 2 deletions
  1. +1
    -1
      ev.h
  2. +5
    -1
      ev_iouring.c

+ 1
- 1
ev.h View File

@ -526,7 +526,7 @@ enum {
EVBACKEND_KQUEUE = 0x00000008U, /* bsd, broken on osx */
EVBACKEND_DEVPOLL = 0x00000010U, /* solaris 8 */ /* NYI */
EVBACKEND_PORT = 0x00000020U, /* solaris 10 */
EVBACKEND_LINUXAIO = 0x00000040U, /* linuix AIO, 4.19+ */
EVBACKEND_LINUXAIO = 0x00000040U, /* linux AIO, 4.19+ */
EVBACKEND_IOURING = 0x00000080U, /* linux io_uring, 5.1+ */
EVBACKEND_ALL = 0x000000FFU, /* all known backends */
EVBACKEND_MASK = 0x0000FFFFU /* all future backends */


+ 5
- 1
ev_iouring.c View File

@ -49,8 +49,9 @@
* e) why 3 mmaps instead of one? one would be more space-efficient,
* and I can't see what benefit three would have (other than being
* somehow resizable/relocatable, but that's apparently not possible).
* (FIXME: newer kernels can use 2 mmaps only, need to look into this).
* f) hmm, it's practiclaly undebuggable (gdb can't access the memory, and
the bizarre way structure offsets are commuinicated makes it hard to
* the bizarre way structure offsets are communicated makes it hard to
* just print the ring buffer heads, even *iff* the memory were visible
* in gdb. but then, that's also ok, really.
* g) well, you cannot specify a timeout when waiting for events. no,
@ -60,8 +61,10 @@
* like a µ-optimisation by the io_uring author for his personal
* applications, to the detriment of everybody else who just wants
* an event loop. but, umm, ok, if that's all, it could be worse.
* (FIXME: jens mentioned timeout commands, need to investigate)
* h) there is a hardcoded limit of 4096 outstanding events. okay,
* at least there is no arbitrary low system-wide limit...
* (FIXME: apparently, this was increased to 32768 in later kernels(
* i) unlike linux aio, you *can* register more then the limit
* of fd events, and the kernel will "gracefully" signal an
* overflow, after which you could destroy and recreate the kernel
@ -69,6 +72,7 @@
* totally insane, but kind of questions the point a high
* performance I/O framework when it doesn't really work
* under stress.
* (FIXME: iouring should no longer drop events, need to investigate)
* j) but, oh my! is has exactly the same bugs as the linux aio backend,
* where some undocumented poll combinations just fail.
* so we need epoll AGAIN as a fallback. AGAIN! epoll!! and of course,


Loading…
Cancel
Save