On Sun, Feb 9, 2014 at 12:44 AM, David Woodhouse <dwmw2 at infradead.org> wrote: > On Sat, 2014-02-08 at 23:10 +0800, Antonio Borneo wrote: >> >> Actually proxy_write() is used also for regular files >> in run_csd_script(): >> fd = mkstemp(fname); >> ... >> ret = proxy_write(vpninfo, fd, (void *)buf, buflen); > > > Hm, do we actually need to use proxy_write() here? Adam did that in > commit cc09a51eb ? quite possibly at my instigation. > > In *theory* we should loop on write() to regular files because it can > return having written fewer bytes than we asked. But in practice it will > never happen, and most userspace would break if it did. > > Still, while it was essentially "free" to do the right thing, it made a > certain amount of sense to be anal about it and to use proxy_write(). > > But once we have to start jumping through hoops to make proxy_write() > choose whether to do write() or send(), perhaps the better answer is > just to use a simple write() in run_csd_script() to write to the file? Agree! Also, the issue I found is the only (ab)use of proxy_write() to write a file. It makes sense to fix it there. New patch in attachment. I don't loop on write() but simply check if it writes fewer bytes than expected. Antonio -------------- next part -------------- A non-text attachment was scrubbed... Name: 0001-fix-bug-in-run_csd_script.patch Type: text/x-patch Size: 1439 bytes Desc: not available URL: <http://lists.infradead.org/pipermail/openconnect-devel/attachments/20140209/e6009e8b/attachment.bin>
