Hi,
Am really sorry about the false alarm - apparently doing
"mke2fs -c -c " on a USB disk on the same hub does very bad things to the throughput of USB wireless device, or rtl8187 specifically. Normal disk usages (copying large files) has no effect at all, so I guess we'll just have to avoid mke2fs or possibly also e2fsck too often :-).
Hin-Tak
--- On Sat, 20/12/08, Hin-Tak Leung <[email protected]> wrote:
> From: Hin-Tak Leung <[email protected]>
> Subject: throughput problem with rtl8187B within last few days (again)?
> To: [email protected], "Larry Finger" <[email protected]>, [email protected]
> Date: Saturday, 20 December, 2008, 1:32 AM
> Hi,
>
> I am using v2.6.28-rc8-8109-g6ec77a1 (about 3 days old,
> 2008-12-17/12-18 thereabouts) and transferring some files
> across and the throughput is back to about 50KB/s again -
> should be about 10x - and iwconfig stays at 1Mbit; doing
> iwconfig rate other numbers doesn't change anything.
>
> I am tracking wireless testing quite regularly so if it
> breaks, the breakage hasn't been more than a week, I
> think; one thing I am not sure about is that I am running
> "mke2fs -c -c " on a USB disk on the same hub (it
> is a laptop so it is a "logical" hub) at the
> moment. So I am wondering if anything breaks in the last few
> days, or other USB activities can seriously affect the
> driver.
>
> I'll try again and write again when the mke2fs -c -c
> finishes (about 30 hours for 250GB). Meanwhile, either of
> you want to say anything?
>
> Hin-Tak
--- On Sat, 20/12/08, Larry Finger <[email protected]> wrote:
> Hin-Tak Leung wrote:
> > Hi,
> >
> > Am really sorry about the false alarm - apparently
> doing
> > "mke2fs -c -c " on a USB disk on the same
> hub does very bad things to the throughput of USB wireless
> device, or rtl8187 specifically. Normal disk usages (copying
> large files) has no effect at all, so I guess we'll just
> have to avoid mke2fs or possibly also e2fsck too often :-).
>
> It must do a lot of reads/writes and occupies both USB
> bandwidth and the USB
> queues. I've never even formatted a USB disk. Having
> more than one hub has
> advantages.
Yes, must be; However, it is interesting that normal usage (copying files between two partitions on the new disk) does not interfere with the wireless driver, compared to doing mke2fs/badblock/e2fsck .
My rtl8187B is an integrated part of a (toshiba) laptop - interesting enough lsusb shows 6 buses (one USB 2.0 root hub and 5 USB 1.1 root hubs),
but all 4 external USB sockets are wired up to the that one USB 2.0 root hub for which the internal rtl8187B is also on.
> Sorry I didn't respond earlier. Gmail decided your
> first post was spam. It was
> the only false positive of about 500 in the Spam list.
Maybe because it was coming from *@sourceforge.net .
Hin-Tak Leung wrote:
> Hi,
>
> Am really sorry about the false alarm - apparently doing
> "mke2fs -c -c " on a USB disk on the same hub does very bad things to the throughput of USB wireless device, or rtl8187 specifically. Normal disk usages (copying large files) has no effect at all, so I guess we'll just have to avoid mke2fs or possibly also e2fsck too often :-).
It must do a lot of reads/writes and occupies both USB bandwidth and the USB
queues. I've never even formatted a USB disk. Having more than one hub has
advantages.
Sorry I didn't respond earlier. Gmail decided your first post was spam. It was
the only false positive of about 500 in the Spam list.
Larry