hashsplit: use posix_fadvise(DONTNEED) when available.

When reading through large disk images to back them up, we'll only end up
reading the data once, but it still takes up space in the kernel's disk
cache.  If you're backing up a whole disk full of stuff, that's bad news for
anything else running on your system, which will rapidly have its stuff
dumped out of cache to store a bunch of stuff bup will never look at again.

The posix_fadvise() call actually lets us tell the kernel we won't be using
this data anymore, thus greatly reducing our hit on the disk cache.

Theoretically it improves things, anyway.  I haven't been able to come up
with a really scientific way to test it, since of course *bup's* performance
is expected to be the same either way (we're only throwing away stuff we're
done using).  It really does throw things out of cache, though, so the rest
follows logically at least.

diff --git a/lib/bup/_hashsplit.c b/lib/bup/_hashsplit.c
index c8a3cde1b112100d36896b049e7cca02cce71080..937b3ad578e20a654b2d21cfde970222cc5ceedf 100644 (file)
--- a/lib/bup/_hashsplit.c
+++ b/lib/bup/_hashsplit.c
@@ -163,6 +163,19 @@ static PyObject *open_noatime(PyObject *self, PyObject *args)
 }
 
 
+static PyObject *fadvise_done(PyObject *self, PyObject *args)
+{
+    int fd = -1;
+    long long ofs = 0;
+    if (!PyArg_ParseTuple(args, "iL", &fd, &ofs))
+       return NULL;
+#ifdef POSIX_FADV_DONTNEED
+    posix_fadvise(fd, 0, ofs, POSIX_FADV_DONTNEED);
+#endif    
+    return Py_BuildValue("");
+}
+
+
 static PyMethodDef hashsplit_methods[] = {
     { "blobbits", blobbits, METH_VARARGS,
        "Return the number of bits in the rolling checksum." },
@@ -174,6 +187,8 @@ static PyMethodDef hashsplit_methods[] = {
        "Write random bytes to the given file descriptor" },
     { "open_noatime", open_noatime, METH_VARARGS,
        "open() the given filename for read with O_NOATIME if possible" },
+    { "fadvise_done", fadvise_done, METH_VARARGS,
+       "Inform the kernel that we're finished with earlier parts of a file" },
     { NULL, NULL, 0, NULL },  // sentinel
 };
 

diff --git a/lib/bup/hashsplit.py b/lib/bup/hashsplit.py
index 764ff55f0603c842ac83da2cef2d00b46496c28c..03232b4f244ee0594591cf0dc6ffe893d6ee71bc 100644 (file)
--- a/lib/bup/hashsplit.py
+++ b/lib/bup/hashsplit.py
@@ -47,9 +47,13 @@ def splitbuf(buf):
 
 def blobiter(files):
     for f in files:
+        ofs = 0
         while 1:
+            fadvise_done(f, max(0, ofs - 1024*1024))
             b = f.read(BLOB_HWM)
+            ofs += len(b)
             if not b:
+                fadvise_done(f, ofs)
                 break
             yield b
 
@@ -168,3 +172,9 @@ def open_noatime(name):
         except:
             pass
         raise
+
+
+def fadvise_done(f, ofs):
+    assert(ofs >= 0)
+    if ofs > 0:
+        _hashsplit.fadvise_done(f.fileno(), ofs)