IRC.

1 files changed, 39 insertions, 0 deletions

diff --git a/hurd/translator/ext2fs/internal_allocator.mdwn b/hurd/translator/ext2fs/internal_allocator.mdwn
new file mode 100644
index 00000000..f3678a28
--- /dev/null
+++ b/hurd/translator/ext2fs/internal_allocator.mdwn
@@ -0,0 +1,39 @@
+[[!meta copyright="Copyright © 2012 Free Software Foundation, Inc."]]
+
+[[!meta license="""[[!toggle id="license" text="GFDL 1.2+"]][[!toggleable
+id="license" text="Permission is granted to copy, distribute and/or modify this
+document under the terms of the GNU Free Documentation License, Version 1.2 or
+any later version published by the Free Software Foundation; with no Invariant
+Sections, no Front-Cover Texts, and no Back-Cover Texts.  A copy of the license
+is included in the section entitled [[GNU Free Documentation
+License|/fdl]]."]]"""]]
+
+[[!tag open_issue_documentation]]
+
+
+# IRC, freenode, #hurd, 2012-07-30
+
+    <mcsim> Why for big buffers in ext2fs used own allocator, that just
+      allocates many pages at once, instead of using malloc?
+    <mcsim> i.e. can I replace it with malloc, because it just complicates
+      things?
+    <braunr> mcsim: probably because of alignment
+    <braunr> what gets complicated by that ?
+    <mcsim> braunr: than valloc?
+    <mcsim> braunr: this allocator allows to allocate only buffer with size of
+      vm_page_size.
+    <mcsim> valloc just would be clearer.
+    <braunr> valloc ?
+    <braunr> valloc is obsolete
+    <mcsim> braunr: than memalign or posix_memalign?
+    <mcsim> memalign obsolete too... would posix_memalign be eligible?
+    <braunr> mcsim: why memalign instead of the custom allocator ?
+    <mcsim> because, I think, it is clearer. Also, since I need to allocate any
+      amount of pages, not just one, I have to edit custom allocator. Although
+      it is not hard, but using ready stuff seems more sane for me.
+    <mcsim> braunr: ^
+    <braunr> right, but make sure posix_memalign doesn't create too much
+      overhead
+    <mcsim> braunr: what kind of overhead?
+    <braunr> fragmentation
+    <braunr> i assume the glibc implementation is careful about that, but still