switch io_fds from array to newly implemented (hopefully thread-safe) iarray

switch epoll from level triggering to edge triggering

CHANGES

@@ -6,12 +6,14 @@
   change buffer to have a destructor function pointer
   SECURITY: fmt_strn would write one byte too many (returned right length though, so usually not a problem as that byte would be overwritten with \0 by the caller anyway)
   fmt_pad and fmt_fill fail more gracefully when srclen > maxlen
-  You can now say $ make WERROR=-Werror (compiling t.c will fail but that's not part of the library)
+  You can now say $ make WERROR=-Werror (compiling t.c may fail but that's strictly speaking not part of the library)
   scan_html now decodes HTML entities based on the actual w3c list (from entities.json, say gmake update to fetch the current version)
   added fmt_escapechar* to fmt.h (implement various escaping mechanisms also found in textcode but for a single char not a whole string, and they always escape, not just when they think it's needed)
   scan_ushort was supposed to abort early and return 5 when attempting to parse "65536", because the result does not fit.  It did not.  Now it does.
   scan_*long, scan_*int, scan_*short now properly abort if the number would not fit
   SECURITY: check for integer overflow in stralloc_ready
+  switch io_fds from array to newly implemented (hopefully thread-safe) iarray
+  switch epoll from level triggering to edge triggering
 
 0.29:
   save 8 bytes in taia.h for 64-bit systems

array/iarray_allocate.c

@@ -17,7 +17,7 @@ static iarray_page* new_page(size_t pagesize) {
 }
 
 void* iarray_allocate(iarray* ia,size_t pos) {
-  size_t index;
+  size_t index,prevlen=ia->len;
   iarray_page** p=&ia->pages[pos%(sizeof(ia->pages)/sizeof(ia->pages[0]))];
   iarray_page* newpage=0;
   for (index=0; pos<index+ia->elemperpage; index+=ia->elemperpage) {
@@ -28,7 +28,11 @@ void* iarray_allocate(iarray* ia,size_t pos) {
 	newpage=0;
     }
     if (index+ia->elemperpage>pos) {
+      size_t l;
       if (newpage) munmap(newpage,ia->bytesperpage);
+      do {
+	l=__CAS(&ia->len,prevlen,pos);
+      } while (l<pos);
       return &(*p)->data[(pos-index)*ia->elemsize];
     }
     p=&(*p)->next;

buffer/buffer_mmapread.c

@@ -6,7 +6,7 @@ static ssize_t op() {
 }
 
 int buffer_mmapread(buffer* b,const char* filename) {
-  if (!(b->x=mmap_read(filename,&b->n))) return -1;
+  if (!(b->x=(char*)mmap_read(filename,&b->n))) return -1;
   b->p=0; b->a=b->n;
   b->fd=-1;
   b->op=op;

iarray.h

@@ -24,12 +24,13 @@ typedef struct _iarray_page {
 
 typedef struct {
   iarray_page* pages[16];
-  size_t elemsize,elemperpage,bytesperpage;
+  size_t elemsize,elemperpage,bytesperpage,len;
 } iarray;
 
 void iarray_init(iarray* ia,size_t elemsize);
 void* iarray_get(iarray* ia,size_t pos);
 void* iarray_allocate(iarray* ia,size_t pos);
+size_t iarray_length(iarray* ia);
 
 /* WARNING: do not use the array during or after iarray_free, make sure
  * no threads are potentially doing anything with the iarray while it is

io/io_block.c

@@ -13,7 +13,7 @@
 #endif
 
 void io_block(int64 d) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
 #ifdef __MINGW32__
   unsigned long i=0;
   if (ioctlsocket( d, FIONBIO, &i)==0)

io/io_canread.c

@@ -14,7 +14,7 @@ int64 io_canread() {
   if (first_readable==-1)
 #ifdef HAVE_SIGIO
   {
-    if (alt_firstread>=0 && (e=array_get(&io_fds,sizeof(io_entry),alt_firstread)) && e->canread) {
+    if (alt_firstread>=0 && (e=iarray_get(&io_fds,alt_firstread)) && e->canread) {
       debug_printf(("io_canread: normal read queue is empty, swapping in alt read queue (starting with %ld)\n",alt_firstread));
       first_readable=alt_firstread;
       alt_firstread=-1;
@@ -26,7 +26,7 @@ int64 io_canread() {
 #endif
   for (;;) {
     int64 r;
-    e=array_get(&io_fds,sizeof(io_entry),first_readable);
+    e=iarray_get(&io_fds,first_readable);
     if (!e) break;
     r=first_readable;
     first_readable=e->next_read;

io/io_canwrite.c

@@ -10,7 +10,7 @@ int64 io_canwrite() {
   if (first_writeable==-1)
 #ifdef HAVE_SIGIO
   {
-    if (alt_firstwrite>=0 && (e=array_get(&io_fds,sizeof(io_entry),alt_firstwrite)) && e->canwrite) {
+    if (alt_firstwrite>=0 && (e=iarray_get(&io_fds,alt_firstwrite)) && e->canwrite) {
       debug_printf(("io_canwrite: normal write queue is empty, swapping in alt write queue (starting with %ld)\n",alt_firstwrite));
       first_writeable=alt_firstwrite;
       alt_firstwrite=-1;
@@ -22,7 +22,7 @@ int64 io_canwrite() {
 #endif
   for (;;) {
     int64 r;
-    e=array_get(&io_fds,sizeof(io_entry),first_writeable);
+    e=iarray_get(&io_fds,first_writeable);
     if (!e) break;
     r=first_writeable;
     first_writeable=e->next_write;

io/io_close.c

@@ -12,7 +12,7 @@ extern void io_dontwantwrite_really(int64 d,io_entry* e);
 
 void io_close(int64 d) {
   io_entry* e;
-  if ((e=array_get(&io_fds,sizeof(io_entry),d))) {
+  if ((e=iarray_get(&io_fds,d))) {
     e->inuse=0;
     e->cookie=0;
     if (e->kernelwantread) io_dontwantread_really(d,e);

io/io_dontwantread.c

@@ -26,19 +26,10 @@
 
 void io_dontwantread_really(int64 d, io_entry* e) {
   int newfd;
+  (void)d;
   assert(e->kernelwantread);
   newfd=!e->kernelwantwrite;
   io_wanted_fds-=newfd;
-#ifdef HAVE_EPOLL
-  if (io_waitmode==EPOLL) {
-    struct epoll_event x;
-    byte_zero(&x,sizeof(x));	// to shut up valgrind
-    x.events=0;
-    if (e->kernelwantwrite) x.events|=EPOLLOUT;
-    x.data.fd=d;
-    epoll_ctl(io_master,e->kernelwantwrite?EPOLL_CTL_MOD:EPOLL_CTL_DEL,d,&x);
-  }
-#endif
 #ifdef HAVE_KQUEUE
   if (io_waitmode==KQUEUE) {
     struct kevent kev;
@@ -63,7 +54,7 @@ void io_dontwantread_really(int64 d, io_entry* e) {
 }
 
 void io_dontwantread(int64 d) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (e) {
     if (e->canread)
       io_dontwantread_really(d,e);

io/io_dontwantwrite.c

@@ -32,19 +32,10 @@
 
 void io_dontwantwrite_really(int64 d,io_entry* e) {
   int newfd;
+  (void)d;
   assert(e->kernelwantwrite);
   newfd=!e->kernelwantread;
   io_wanted_fds-=newfd;
-#ifdef HAVE_EPOLL
-  if (io_waitmode==EPOLL) {
-    struct epoll_event x;
-    byte_zero(&x,sizeof(x));	// to shut up valgrind
-    x.events=0;
-    if (e->wantread) x.events|=EPOLLIN;
-    x.data.fd=d;
-    epoll_ctl(io_master,e->kernelwantread?EPOLL_CTL_MOD:EPOLL_CTL_DEL,d,&x);
-  }
-#endif
 #ifdef HAVE_KQUEUE
   if (io_waitmode==KQUEUE) {
     struct kevent kev;
@@ -69,7 +60,7 @@ void io_dontwantwrite_really(int64 d,io_entry* e) {
 }
 
 void io_dontwantwrite(int64 d) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (e) {
     if (e->canwrite)
       io_dontwantwrite_really(d,e);

io/io_eagain.c

@@ -1,7 +1,7 @@
 #include "io_internal.h"
 
 void io_eagain(int64 d) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (e) {
     if (e->wantread) e->canread=0;
     if (e->wantwrite) e->canwrite=0;

io/io_fd.c

@@ -25,11 +25,18 @@
 #include <sys/devpoll.h>
 #endif
 
+#ifdef __dietlibc__
+#include <sys/atomic.h>
+#else
+#define __CAS(ptr,oldval,newval) __sync_val_compare_and_swap(ptr,oldval,newval)
+#endif
+
 #ifdef __MINGW32__
 #include <stdio.h>
 extern HANDLE io_comport;
 #endif
-array io_fds;
+iarray io_fds;
+static int io_fds_inited;
 uint64 io_wanted_fds;
 array io_pollfds;
 enum __io_waitmode io_waitmode;
@@ -52,7 +59,17 @@ static io_entry* io_fd_internal(int64 d) {
   if ((r=fcntl(d,F_GETFL,0)) == -1)
     return 0;	/* file descriptor not open */
 #endif
-  if (!(e=array_allocate(&io_fds,sizeof(io_entry),d))) return 0;
+  /* Problem: we might be the first to use io_fds. We need to make sure
+   * we are the only ones to initialize it.  So set io_fds_inited to 2
+   * and not to 1.  We know we are done when it is 1.  We know we need
+   * to do something when it is 0.  We know somebody else is doing it
+   * when it is 2. */
+  if (__CAS(&io_fds_inited,0,2)==0) {
+    iarray_init(&io_fds,sizeof(io_entry));
+    io_fds_inited=1;
+  } else
+    do { asm("" : : : "memory"); } while (io_fds_inited!=1);
+  if (!(e=iarray_allocate(&io_fds,d))) return 0;
   if (e->inuse) return e;
   byte_zero(e,sizeof(io_entry));
   e->inuse=1;

io/io_finishandshutdown.c

@@ -1,6 +1,6 @@
 #include "io_internal.h"
 
 void io_finishandshutdown(void) {
-  array_reset(&io_fds);
+  iarray_free(&io_fds);
   array_reset(&io_pollfds);
 }

io/io_getcookie.c

@@ -3,6 +3,6 @@
 
 void* io_getcookie(int64 d) {
   io_entry* e;
-  e=array_get(&io_fds,sizeof(io_entry),d);
+  e=iarray_get(&io_fds,d);
   return e?e->cookie:0;
 }

io/io_mmapwritefile.c

@@ -15,7 +15,7 @@ int64 io_mmapwritefile(int64 out,int64 in,uint64 off,uint64 bytes,io_write_callb
   char buf[BUFSIZE];
   int n,m;
   uint64 sent=0;
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),out);
+  io_entry* e=iarray_get(&io_fds,out);
   if (e) {
     const char* c;
     unsigned long left;

io/io_nonblock.c

@@ -13,7 +13,7 @@
 #endif
 
 void io_nonblock(int64 d) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
 #ifdef __MINGW32__
   unsigned long i=1;
   if (ioctlsocket( d, FIONBIO, &i)==0)

io/io_sendfile.c

@@ -86,7 +86,7 @@ _syscall4(int,sendfile,int,out,int,in,long *,offset,unsigned long,count)
 
 int64 io_sendfile(int64 s,int64 fd,uint64 off,uint64 n) {
   off_t o=off;
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),s);
+  io_entry* e=iarray_get(&io_fds,s);
   off_t i;
   uint64 done=0;
   /* What a spectacularly broken design for sendfile64.
@@ -118,7 +118,7 @@ int64 io_sendfile(int64 s,int64 fd,uint64 off,uint64 n) {
 #include <mswsock.h>
 
 int64 io_sendfile(int64 out,int64 in,uint64 off,uint64 bytes) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),out);
+  io_entry* e=iarray_get(&io_fds,out);
   if (!e) { errno=EBADF; return -3; }
   if (e->sendfilequeued==1) {
     /* we called TransmitFile, and it returned. */

io/io_setcookie.c

@@ -4,6 +4,6 @@
 
 void io_setcookie(int64 d,void* cookie) {
   io_entry* e;
-  if ((e=array_get(&io_fds,sizeof(io_entry),d)))
+  if ((e=iarray_get(&io_fds,d)))
     e->cookie=cookie;
 }

io/io_timeout.c

@@ -1,7 +1,7 @@
 #include "io_internal.h"
 
 void io_timeout(int64 d,tai6464 t) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (e)
     e->timeout=t;
 }

io/io_timeouted.c

@@ -2,13 +2,13 @@
 
 int64 io_timeouted() {
   tai6464 now;
-  static long ptr;
+  static size_t ptr;
   io_entry* e;
-  long alen=array_length(&io_fds,sizeof(io_entry));
+  size_t alen=iarray_length(&io_fds);
   taia_now(&now);
   ++ptr;
   if (ptr>=alen) ptr=0;
-  e=array_get(&io_fds,sizeof(io_entry),ptr);
+  e=iarray_get(&io_fds,ptr);
   if (!e) return -1;
   for (;ptr<alen; ++ptr,++e) {
     if (e->inuse && e->timeout.sec.x && taia_less(&e->timeout,&now)) {

io/io_tryread.c

@@ -20,7 +20,7 @@
  * we are called. */
 
 int64 io_tryread(int64 d,char* buf,int64 len) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (!e) { errno=EBADF; return -3; }
   if (len<0) { errno=EINVAL; return -3; }
   if (e->readqueued==2) {
@@ -82,7 +82,7 @@ int64 io_tryread(int64 d,char* buf,int64 len) {
   long r;
   struct itimerval old,new;
   struct pollfd p;
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (!e) { errno=EBADF; return -3; }
   if (!e->nonblock) {
     p.fd=d;

io/io_tryreadtimeout.c

@@ -5,7 +5,7 @@ int64 io_tryreadtimeout(int64 d,char* buf,int64 len) {
   int64 r=io_tryread(d,buf,len);
   if (r==-1) {
     tai6464 x;
-    io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+    io_entry* e=iarray_get(&io_fds,d);
     taia_now(&x);
     if (!taia_less(&x,&e->timeout)) {
       errno=ETIMEDOUT;

io/io_trywrite.c

@@ -19,7 +19,7 @@
  * stuff on I/O batches. */
 
 int64 io_trywrite(int64 d,const char* buf,int64 len) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   int r;
   if (!e) { errno=EBADF; return -3; }
   if (!e->nonblock) {
@@ -75,7 +75,7 @@ int64 io_trywrite(int64 d,const char* buf,int64 len) {
   long r;
   struct itimerval old,new;
   struct pollfd p;
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   io_sigpipe();
   if (!e) { errno=EBADF; return -3; }
   if (!e->nonblock) {

io/io_trywritetimeout.c

@@ -5,7 +5,7 @@ int64 io_trywritetimeout(int64 d,const char* buf,int64 len) {
   int64 r=io_trywrite(d,buf,len);
   if (r==-1) {
     tai6464 x;
-    io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+    io_entry* e=iarray_get(&io_fds,d);
     taia_now(&x);
     if (!taia_less(&x,&e->timeout)) {
       errno=ETIMEDOUT;

io/io_waitread.c

@@ -11,7 +11,7 @@
 
 int64 io_waitread(int64 d,char* buf,int64 len) {
   long r;
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (!e) { errno=EBADF; return -3; }
   if (e->nonblock) {
     unsigned long i=0;
@@ -32,7 +32,7 @@ int64 io_waitread(int64 d,char* buf,int64 len) {
 int64 io_waitread(int64 d,char* buf,int64 len) {
   long r;
   struct pollfd p;
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (!e) { errno=EBADF; return -3; }
   if (e->nonblock) {
 again:

io/io_waituntil2.c

@@ -28,6 +28,11 @@
 #include <sys/devpoll.h>
 #endif
 
+#ifdef __dietlibc__
+#include <fmt.h>
+#include <write12.h>
+#endif
+
 #ifdef DEBUG
 #include <stdio.h>
 #endif
@@ -119,97 +124,38 @@ int64 io_waituntil2(int64 milliseconds) {
   if (io_waitmode==EPOLL) {
     int n;
     struct epoll_event y[100];
+    io_entry* e;
+    if (alt_firstread>=0 && (e=iarray_get(&io_fds,alt_firstread)) && e->canread) return 1;
+    if (alt_firstwrite>=0 && (e=iarray_get(&io_fds,alt_firstwrite)) && e->canwrite) return 1;
     if ((n=epoll_wait(io_master,y,100,milliseconds))==-1) return -1;
-    for (i=n-1; i>=0; --i) {
+    for (i=0; i<n; ++i) {
-      io_entry* e=array_get(&io_fds,sizeof(io_entry),y[i].data.fd);
+      e=iarray_get(&io_fds,y[i].data.fd);
       if (e) {
-	int curevents=0,newevents;
+	if (y[i].events&(POLLERR|POLLHUP)) {
-	if (e->kernelwantread) curevents |= EPOLLIN;
-	if (e->kernelwantwrite) curevents |= EPOLLOUT;
-
-#ifdef DEBUG
-	if ((y[i].events&(EPOLLIN|EPOLLPRI|EPOLLRDNORM|EPOLLRDBAND)) && !e->kernelwantread)
-	  printf("got unexpected read event on fd #%d\n",y[i].data.fd);
-	if ((y[i].events&EPOLLOUT) && !e->kernelwantwrite)
-	  printf("got unexpected write event on fd #%d\n",y[i].data.fd);
-#endif
-
-	if (y[i].events&(EPOLLERR|EPOLLHUP)) {
 	  /* error; signal whatever app is looking for */
-	  if (e->wantread) y[i].events|=EPOLLIN;
+	  if (e->wantread) y[i].events|=POLLIN;
-	  if (e->wantwrite) y[i].events|=EPOLLOUT;
+	  if (e->wantwrite) y[i].events|=POLLOUT;
 	}
-
+	if (y[i].events&POLLIN && !e->canread) {
-	newevents=0;
+	  debug_printf(("io_waituntil2: enqueueing %ld in normal read queue before %ld\n",info.si_fd,first_readable));
-	if (!e->canread || e->wantread) {
-	  newevents|=EPOLLIN;
-	  e->kernelwantread=1;
-	} else
-	  e->kernelwantread=0;
-	if (!e->canwrite || e->wantwrite) {
-	  newevents|=EPOLLOUT;
-	  e->kernelwantwrite=1;
-	} else
-	  e->kernelwantwrite=0;
-
-	/* if we think we can not read, but the kernel tells us that we
-	 * can, put this fd in the relevant data structures */
-	if (!e->canread && (y[i].events&(EPOLLIN|EPOLLPRI|EPOLLRDNORM|EPOLLRDBAND))) {
-	  if (e->canread) {
-	    newevents &= ~EPOLLIN;
-	  } else {
 	  e->canread=1;
-	    if (e->wantread) {
 	  e->next_read=first_readable;
 	  first_readable=y[i].data.fd;
 	}
-	  }
+	if (y[i].events&POLLOUT && !e->canwrite) {
-	}
+	  debug_printf(("io_waituntil2: enqueueing %ld in normal write queue before %ld\n",info.si_fd,first_writeable));
-	/* if the kernel says the fd is writable, ... */
-	if (y[i].events&EPOLLOUT) {
-	  /* Usually, if the kernel says a descriptor is writable, we
-	   * note it and do not tell the kernel not to tell us again.
-	   * The idea is that once we notify the caller that the fd is
-	   * writable, and the caller handles the event, the caller will
-	   * just ask to be notified of future write events again.  We
-	   * are trying to save the superfluous epoll_ctl syscalls.
-	   * If e->canwrite is set, then this gamble did not work out.
-	   * We told the caller, yet after the caller is done we still
-	   * got another write event.  Clearly the user is implementing
-	   * some kind of throttling and we can tell the kernel to leave
-	   * us alone for now. */
-	  if (e->canwrite) {
-	    newevents &= ~EPOLLOUT;
-	    e->kernelwantwrite=0;
-	  } else {
-	    /* If !e->wantwrite: The laziness optimization in
-	     * io_dontwantwrite hit.  We did not tell the kernel that we
-	     * are no longer interested in writing to save the syscall.
-	     * Now we know we could write if we wanted; remember that
-	     * and then go on. */
 	  e->canwrite=1;
-	    if (e->wantwrite) {
 	  e->next_write=first_writeable;
 	  first_writeable=y[i].data.fd;
 	}
-	  }
-	}
-
-	if (newevents != curevents) {
-#if 0
-	  printf("canread %d, wantread %d, kernelwantread %d, canwrite %d, wantwrite %d, kernelwantwrite %d\n",
-		 e->canread,e->wantread,e->kernelwantread,e->canwrite,e->wantwrite,e->kernelwantwrite);
-	  printf("newevents: read %d write %d\n",!!(newevents&EPOLLIN),!!(newevents&EPOLLOUT));
-#endif
-	  y[i].events=newevents;
-	  if (newevents) {
-	    epoll_ctl(io_master,EPOLL_CTL_MOD,y[i].data.fd,y+i);
-	  } else {
-	    epoll_ctl(io_master,EPOLL_CTL_DEL,y[i].data.fd,y+i);
-	    --io_wanted_fds;
-	  }
-	}
       } else {
+#ifdef __dietlibc__
+	char buf[FMT_ULONG];
+	buf[fmt_ulong(buf,y[i].data.fd)]=0;
+	__write2("got epoll event on invalid fd ");
+	__write2(buf);
+	__write2("!\n");
+#endif
 	epoll_ctl(io_master,EPOLL_CTL_DEL,y[i].data.fd,y+i);
       }
     }
@@ -224,7 +170,7 @@ int64 io_waituntil2(int64 milliseconds) {
     ts.tv_sec=milliseconds/1000; ts.tv_nsec=(milliseconds%1000)*1000000;
     if ((n=kevent(io_master,0,0,y,100,milliseconds!=-1?&ts:0))==-1) return -1;
     for (i=n-1; i>=0; --i) {
-      io_entry* e=array_get(&io_fds,sizeof(io_entry),y[--n].ident);
+      io_entry* e=iarray_get(&io_fds,y[--n].ident);
       if (e) {
 	if (y[n].flags&EV_ERROR) {
 	  /* error; signal whatever app is looking for */
@@ -260,7 +206,7 @@ int64 io_waituntil2(int64 milliseconds) {
     timeout.dp_fds=y;
     if ((n=ioctl(io_master,DP_POLL,&timeout))==-1) return -1;
     for (i=n-1; i>=0; --i) {
-      io_entry* e=array_get(&io_fds,sizeof(io_entry),y[--n].fd);
+      io_entry* e=iarray_get(&io_fds,y[--n].fd);
       if (e) {
 	if (y[n].revents&(POLLERR|POLLHUP|POLLNVAL)) {
 	  /* error; signal whatever app is looking for */
@@ -296,8 +242,8 @@ int64 io_waituntil2(int64 milliseconds) {
     struct timespec ts;
     int r;
     io_entry* e;
-    if (alt_firstread>=0 && (e=array_get(&io_fds,sizeof(io_entry),alt_firstread)) && e->canread) return 1;
+    if (alt_firstread>=0 && (e=iarray_get(&io_fds,alt_firstread)) && e->canread) return 1;
-    if (alt_firstwrite>=0 && (e=array_get(&io_fds,sizeof(io_entry),alt_firstwrite)) && e->canwrite) return 1;
+    if (alt_firstwrite>=0 && (e=iarray_get(&io_fds,alt_firstwrite)) && e->canwrite) return 1;
     if (milliseconds==-1)
       r=sigwaitinfo(&io_ss,&info);
     else {
@@ -311,7 +257,7 @@ int64 io_waituntil2(int64 milliseconds) {
       goto dopoll;
     default:
       if (r==io_signum) {
-	io_entry* e=array_get(&io_fds,sizeof(io_entry),info.si_fd);
+	io_entry* e=iarray_get(&io_fds,info.si_fd);
 	if (e) {
 	  if (info.si_band&(POLLERR|POLLHUP)) {
 	    /* error; signal whatever app is looking for */
@@ -351,7 +297,7 @@ dopoll:
   }
   fprintf(stderr,"Calling GetQueuedCompletionStatus %p...",io_comport);
   if (GetQueuedCompletionStatus(io_comport,&numberofbytes,&x,&o,milliseconds==-1?INFINITE:milliseconds)) {
-    io_entry* e=array_get(&io_fds,sizeof(io_entry),x);
+    io_entry* e=iarray_get(&io_fds,x);
     fprintf(stderr," OK.  Got %x, e=%p\n",x,e);
     if (!e) return 0;
     e->errorcode=0;
@@ -398,7 +344,7 @@ dopoll:
     /* we got a completion packet for a failed I/O operation */
     err=GetLastError();
     if (err==WAIT_TIMEOUT) return 0;	/* or maybe not */
-    e=array_get(&io_fds,sizeof(io_entry),x);
+    e=iarray_get(&io_fds,x);
     if (!e) return 0;	/* WTF?! */
     e->errorcode=err;
     if (o==&e->or && (e->readqueued || e->acceptqueued)) {
@@ -422,8 +368,8 @@ dopoll:
     return 1;
   }
 #else
-  for (i=r=0; i<array_length(&io_fds,sizeof(io_entry)); ++i) {
+  for (i=r=0; (size_t)i<iarray_length(&io_fds); ++i) {
-    io_entry* e=array_get(&io_fds,sizeof(io_entry),i);
+    io_entry* e=iarray_get(&io_fds,i);
     if (!e) return -1;
     e->canread=e->canwrite=0;
     if (e->wantread || e->wantwrite) {
@@ -439,7 +385,7 @@ dopoll:
   p=array_start(&io_pollfds);
   if ((i=poll(array_start(&io_pollfds),r,milliseconds))<1) return -1;
   for (j=r-1; j>=0; --j) {
-    io_entry* e=array_get(&io_fds,sizeof(io_entry),p->fd);
+    io_entry* e=iarray_get(&io_fds,p->fd);
     if (p->revents&(POLLERR|POLLHUP|POLLNVAL)) {
       /* error; signal whatever app is looking for */
       if (e->wantread) p->revents|=POLLIN;

io/io_waitwrite.c

@@ -11,7 +11,7 @@
 
 int64 io_waitwrite(int64 d,const char* buf,int64 len) {
   long r;
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (!e) { errno=EBADF; return -3; }
   if (e->nonblock) {
     unsigned long i=0;
@@ -32,7 +32,7 @@ int64 io_waitwrite(int64 d,const char* buf,int64 len) {
 int64 io_waitwrite(int64 d,const char* buf,int64 len) {
   long r;
   struct pollfd p;
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   io_sigpipe();
   if (!e) { errno=EBADF; return -3; }
   if (e->nonblock) {

io/io_wantread.c

@@ -35,16 +35,6 @@ void io_wantread_really(int64 d,io_entry* e) {
   assert(!e->kernelwantread);
   newfd=!e->kernelwantwrite;
   io_wanted_fds+=newfd;
-#ifdef HAVE_EPOLL
-  if (io_waitmode==EPOLL) {
-    struct epoll_event x;
-    byte_zero(&x,sizeof(x));	// to shut up valgrind
-    x.events=EPOLLIN;
-    if (e->kernelwantwrite) x.events|=EPOLLOUT;
-    x.data.fd=d;
-    epoll_ctl(io_master,e->kernelwantwrite?EPOLL_CTL_MOD:EPOLL_CTL_ADD,d,&x);
-  }
-#endif
 #ifdef HAVE_KQUEUE
   if (io_waitmode==KQUEUE) {
     struct kevent kev;
@@ -63,16 +53,26 @@ void io_wantread_really(int64 d,io_entry* e) {
     write(io_master,&x,sizeof(x));
   }
 #endif
-#ifdef HAVE_SIGIO
+#if defined(HAVE_SIGIO) || defined(HAVE_EPOLL)
-  if (io_waitmode==_SIGIO) {
+  if (io_waitmode==_SIGIO || io_waitmode==EPOLL) {
     struct pollfd p;
+    if (io_waitmode==EPOLL && !e->epolladded) {
+      struct epoll_event x;
+      byte_zero(&x,sizeof(x));	// shut up valgrind
+      x.events=EPOLLIN|EPOLLOUT|EPOLLET;
+      x.data.fd=d;
+      epoll_ctl(io_master,EPOLL_CTL_ADD,d,&x);
+      e->epolladded=1;
+    }
+    if (e->canread==0) {
       p.fd=d;
       p.events=POLLIN;
       switch (poll(&p,1,0)) {
       case 1: e->canread=1; break;
-    case 0: e->canread=0; break;
+//      case 0: e->canread=0; break;
       case -1: return;
       }
+    }
     if (e->canread) {
       debug_printf(("io_wantread: enqueueing %lld in normal read queue (next is %ld)\n",d,first_readable));
       e->next_read=first_readable;
@@ -113,7 +113,7 @@ queueread:
 }
 
 void io_wantread(int64 d) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (!e || e->wantread) return;
   if (e->canread) {
     e->next_read=first_readable;

io/io_wantwrite.c

@@ -40,16 +40,6 @@ void io_wantwrite_really(int64 d, io_entry* e) {
   assert(!e->kernelwantwrite);	/* we should not be here if we already told the kernel we want to write */
   newfd=(!e->kernelwantread);
   io_wanted_fds+=newfd;
-#ifdef HAVE_EPOLL
-  if (io_waitmode==EPOLL) {
-    struct epoll_event x;
-    byte_zero(&x,sizeof(x));	// to shut up valgrind
-    x.events=EPOLLOUT;
-    if (e->kernelwantread) x.events|=EPOLLIN;
-    x.data.fd=d;
-    epoll_ctl(io_master,e->kernelwantread?EPOLL_CTL_MOD:EPOLL_CTL_ADD,d,&x);
-  }
-#endif
 #ifdef HAVE_KQUEUE
   if (io_waitmode==KQUEUE) {
     struct kevent kev;
@@ -68,16 +58,26 @@ void io_wantwrite_really(int64 d, io_entry* e) {
     write(io_master,&x,sizeof(x));
   }
 #endif
-#ifdef HAVE_SIGIO
+#if defined(HAVE_SIGIO) || defined(HAVE_EPOLL)
-  if (io_waitmode==_SIGIO) {
+  if (io_waitmode==_SIGIO || io_waitmode==EPOLL) {
     struct pollfd p;
+    if (io_waitmode==EPOLL && !e->epolladded) {
+      struct epoll_event x;
+      byte_zero(&x,sizeof(x));	// shut up valgrind
+      x.events=EPOLLIN|EPOLLOUT|EPOLLET;
+      x.data.fd=d;
+      epoll_ctl(io_master,EPOLL_CTL_ADD,d,&x);
+      e->epolladded=1;
+    }
+    if (e->canwrite==0) {
       p.fd=d;
       p.events=POLLOUT;
       switch (poll(&p,1,0)) {
       case 1: e->canwrite=1; break;
-    case 0: e->canwrite=0; break;
+//      case 0: e->canwrite=0; break;
       case -1: return;
       }
+    }
     if (e->canwrite) {
       debug_printf(("io_wantwrite: enqueueing %lld in normal write queue before %ld\n",d,first_readable));
       e->next_write=first_writeable;
@@ -99,7 +99,7 @@ void io_wantwrite_really(int64 d, io_entry* e) {
 }
 
 void io_wantwrite(int64 d) {
-  io_entry* e=array_get(&io_fds,sizeof(io_entry),d);
+  io_entry* e=iarray_get(&io_fds,d);
   if (!e) return;
   if (e->wantwrite && e->kernelwantwrite) return;
   if (e->canwrite) {

io/iob_send.c

@@ -17,7 +17,7 @@ int64 iob_send(int64 s,io_batch* b) {
 
   if (b->bytesleft==0) return 0;
   sent=-1;
-  e=array_get(&io_fds,sizeof(io_entry),s);
+  e=iarray_get(&io_fds,s);
   if (!e) { errno=EBADF; return -3; }
   if (!(x=array_get(&b->b,sizeof(iob_entry),b->next)))
     return -3;		/* can't happen error */

io_internal.h

@@ -4,6 +4,7 @@
 
 #include "io.h"
 #include "array.h"
+#include "iarray.h"
 #ifdef __MINGW32__
 #include "socket.h"
 my_extern HANDLE io_comport;
@@ -36,6 +37,7 @@ my_extern HANDLE io_comport;
 
 typedef struct {
   tai6464 timeout;
+  int fd;
   unsigned int wantread:1;	/* does the app want to read/write? */
   unsigned int wantwrite:1;
   unsigned int canread:1;	/* do we know we can read/write? */
@@ -44,6 +46,7 @@ typedef struct {
   unsigned int inuse:1;		/* internal consistency checking */
   unsigned int kernelwantread:1;	/* did we tell the kernel we want to read/write? */
   unsigned int kernelwantwrite:1;
+  unsigned int epolladded:1;
 #ifdef __MINGW32__
   unsigned int readqueued:2;
   unsigned int writequeued:2;
@@ -71,7 +74,7 @@ typedef struct {
 extern int io_multithreaded;
 extern int io_sockets[2];
 
-my_extern array io_fds;
+my_extern iarray io_fds;
 my_extern uint64 io_wanted_fds;
 my_extern array io_pollfds;
 

t.c

@@ -51,6 +51,7 @@ int main(int argc,char* argv[]) {
   char buf[1024];
   size_t l;
   unsigned char c;
+  (void)writecb;
   printf("%d\n",(c=scan_fromhex('.')));
   (void)argc;
   (void)argv;

test/io5.c

@@ -47,6 +47,7 @@ int main() {
 	  buffer_putulong(buffer_2,n);
 	  buffer_puts(buffer_2,")");
 	  if (io_fd(n)) {
+	    io_nonblock(n);
 	    io_wantread(n);
 	  } else {
 	    buffer_puts(buffer_2,", but io_fd failed.");