ba569s984@sycom.UUCP (Bill Allen) (06/14/91)
Looks like John "JrComm" Radigan was right. Too many HD partitions mean big trouble with modem speeds >9600. I have 3-4 HDs, with 2-3 partitions each. The 6-12 partitions cause random characters to be dropped during text displays at >9600. Even off-line, "ATI5 modem --> terminal" displays, exhibit this problem. Needless to say, file transfers are nothing but errors/resends. I selectively removed partitions and notices the problem slowly goes away somewhere around the 3rd-5th partition mounting. What's the solution. Bringing my partition count down to a minimum still won't help with 5-7 HDs online. Selling HDs isn't a "fix" either. From JrComm docs- John "JrComm" Radigan is convinced this problem is caused by the system needing to check each partition for "disk-change", like floppies. Has anyone seen a tool (like NoClick for floppies) that works on HDs? (If that is what needs to be done.) Why is the system checking for "change" in fixed HDs anyway???? Why are static, "mounted, but not currently in use" partitions using so much of the CPU's resources? Is a patch to l:Disk-Validator needed? Is a patch to Expansion:gvpscsi needed? Is: RM> gvpscsi 14468 ----rwed 03-Aug-90 14:42:50 RM> : v3.7 (17 Jul 1990) Ralph Babel RM> the very latest release? Is v1.95 13Sep90, the latest release of FaaastPrep? I notice it reports the PRIORITY of each partition as 0, but Steve Tibbets "TaskX" shows every partition as PRIORITY 10. Any clues, anyone? Contact SysOp of 1:120/207 313-473-2020. --- FreeFormBBS v0.20 * Origin: Amy ShareWareHQ 2865files/1743users 313-473-2020 (1:120/207)
rbabel@babylon.rmt.sub.org (Ralph Babel) (06/15/91)
In article <ba569s984.7653@sycom.UUCP>, ba569s984@sycom.UUCP (Bill Allen) writes: > Too many HD partitions mean big trouble with modem speeds > >9600. Cannot be said often enough ... > John "JrComm" Radigan is convinced this problem is caused > by the system needing to check each partition for > "disk-change", like floppies. 1.3 FFS doesn't check for medium changes anyway. It's simply the per-process overhead caused by all the filesystems (timer ticks and such). > Why is the system checking for "change" in fixed HDs > anyway???? It doesn't. Well, at least GVP's SCSI driver doesn't. > Why are static, "mounted, but not currently in use" > partitions using so much of the CPU's resources? See above. > Is a patch to l:Disk-Validator needed? Nope. L:Disk-Validator is used to restore the bitmap of OFS disks. The new 2.0 ROM filesystem has a much lower overhead. > Is a patch to Expansion:gvpscsi needed? Nope, there is no checking going on for fixed disks anyway. > Is [gvpscsi V3.7] the very latest release? No, ask your dealer or GVP for an upgrade. It won't make any difference in the reliability of serial transmission, though (it doesn't cause the problem, so it cannot fix it). > I notice it [FaaastPrep] reports the PRIORITY of each > partition as 0, but Steve Tibbets "TaskX" shows every > partition as PRIORITY 10. You are confusing boot priority and task priority. Ralph
rbabel@babylon.rmt.sub.org (Ralph Babel) (06/16/91)
In article <1991Jun15.223714.28874@Sandelman.OCUnix.on.ca>, mcr@Sandelman.OCUnix.on.ca (Michael Richardson) writes: > Is this a Series I or II? i.e. is it a DMA controller? If > it uses polled I/O (likely inside an IPL 7 loop), then no > wonder you are having problems. GVP's SCSI drivers have *NEVER* used polled I/O; the Impact Series-I uses an on-board buffer of 16K (in most cases), i.e. usually one interrupt per 16K. The Series-II requires one interrupt per command (unless an error or disconnect occurs). In no event does the driver use single-byte I/O. And all of GVP's boards use IPL 2, and all actions are deferred to the driver task level, of course. Ralph
bmaple@kessner.denver.co.us (Bob Maple) (06/16/91)
> Looks like John "JrComm" Radigan was right. Too many HD partitions mean big > trouble with modem speeds >9600. Everytime I plan on doing some downloading at 14.4, I always pull up task-x and zap task priorities. (I actualy seek revenge on the trackdisk.device! If you lower anything, lower it!) Just take TaskX and lower all the partitons you AREN'T using to 0 or -1, and bump them back up when you are done. Changes are the priority you saw in FaaastPrep were boot priorities from what it sounds like. -- ........................................................................... : Bob Maple, The Brazilian : "If Milli Vanilli fall in a forrest, does //: : bmaple@nyx.cs.du.edu : someone else make a sound?" _ // : ::::::::::::::::::::::::::::........................................\\//..:
mcr@Sandelman.OCUnix.on.ca (Michael Richardson) (06/16/91)
In article <ba569s984.7653@sycom.UUCP> ba569s984@sycom.UUCP (Bill Allen) writes: >Too many HD partitions mean big trouble with modem speeds >9600. >I have 3-4 HDs, with 2-3 partitions each. The 6-12 partitions cause >random characters to be dropped during text displays at >9600. >Even off-line, "ATI5 modem --> terminal" displays, exhibit this >problem. a) I assume that you have RTS and CTS handshaking enabled on the Amiga, and on the modem (which sounds like an HST) b) You are not running any background tasks, like a cron. >Needless to say, file transfers are nothing but errors/resends. If you are trying to use the internal serial port at >9600 baud, good luck. It does not do RTS/CTS in hardware, requires one interrupt per character, and has a character buffer (if you can it that) To do >9600 you need hardware handshaking (the means that the CTS line is controlled by the shift register, not by the serial.device). This means an ASDG, Checkpoint, or I guess a CA= board (I understand that the CA= board will only guarantee a maximum of 9600 baud on all ports, but that one might be able to go to higher baud rates if you aren't using all the ports) I think the 6552's used on the PDIO boards support hardware flow control in the chip, but I have no specs on those beasts (yet) >John "JrComm" Radigan is convinced this problem is caused >by the system needing to check each partition for "disk-change", >like floppies. Has anyone seen a tool (like NoClick for So? If your system hasn't got the time to do a couple of disk block reads, you aren't going to be able to do any downloading with a streaming protocol. (Zmodem, btw, can be told the size of the receiver's buffer, and will wait after sending that many characters for an ACK. Of course, few implementions actually support this.) >Is a patch to Expansion:gvpscsi needed? >Is: >RM> gvpscsi 14468 ----rwed 03-Aug-90 14:42:50 >RM> : v3.7 (17 Jul 1990) Ralph Babel >RM> >the very latest release? Is this a Series I or II? i.e. is it a DMA controller? If it uses polled I/O (likely inside an IPL 7 loop), then no wonder you are having problems. >I notice it reports the PRIORITY of each partition as 0, but That refers to the BOOT priority. (what order the system looks for places to boot from) >Steve Tibbets "TaskX" shows every partition as PRIORITY 10. File system handlers run at a default TASK priority of 10. This is so that a user task running at the default priority of 0 that won't completely stop all disk activity if it is CPU bound. You might try: 1> ChangeTaskPri 15 1> JrComm and see what happens. -- :!mcr!: | The postmaster never | So much mail, Michael Richardson | resolves twice. | so little time. HOME: mcr@sandelman.ocunix.on.ca Bell: (613) 237-5629 Small Ottawa nodes contact me about joining ocunix.on.ca!
Greg_Bastow@tvbbs.UUCP (Greg Bastow) (06/16/91)
In a message dated Sat 15 Jun 91 09:04, Ba569s984@sycom.uucp (bill Allen)
wrote:
BA> I selectively removed partitions and notices the problem slowly
BA> goes away somewhere around the 3rd-5th partition mounting.
BA> What's the solution. Bringing my partition count down to a
BA> minimum still won't help with 5-7 HDs online. Selling HDs
BA> isn't a "fix" either.
BA> Any clues, anyone? Contact SysOp of 1:120/207 313-473-2020.
Bill,
As noted there is no 'solution' as AmigaDos must scan all partitions and
take a certain amount of processor time away from the system....one possible
solution is to get a secondary Serial card, with a processor of it's own (ie;
the A2232).
I ran into the exact problem before, and I have solved it 100% by both
getting a A2232, and Acceleratoring my system. Even a 020 (SetCpu'd) should
be enough to give you time for the 14.4 modem. The best solution of course
would be a GVP 50Mhz, or even a 3000 - it's only money :-)
-- Via DLG Pro v0.975b
__________ _ _
/////////// \\ // +----------------------------------------------+
// \\ // | Greg_Bastow@tvbbs.wimsey.bc.ca OR 1:153/910 |
// \\ // | 530 Megs Online, Gaming, Usenet,Fido !FILES! |
// \\// +-----+ Fish Disks Online - Official DLG Site |
// unnel \/ ision BBS +----------------------------------------+bmaple@kessner.denver.co.us (Bob Maple) (06/17/91)
> You might try: > > 1> ChangeTaskPri 15 1> JrComm I find that setting JRComm's priority to 1 greatly helps. I run a two-line BBS on C-Net and if I am downloading at 19.2k and someone is on the other line of the BBS even at 2400, I get tons of errors. Hhowever, if I bump JRComm up to 1, everything goes great. -- ........................................................................... : Bob Maple, The Brazilian : "If Milli Vanilli fall in a forrest, does //: : bmaple@nyx.cs.du.edu : someone else make a sound?" _ // : ::::::::::::::::::::::::::::........................................\\//..:
blgardne@javelin.sim.es.com (Blaine Gardner) (06/17/91)
mcr@Sandelman.OCUnix.on.ca (Michael Richardson) writes: >This means an ASDG, Checkpoint, or I guess a CA= board (I understand >that the CA= board will only guarantee a maximum of 9600 baud on all >ports, but that one might be able to go to higher baud rates if you >aren't using all the ports) The A2232 supports up to 19,200 bps simultaneously on all 7 ports. There is also an oddball 100 Kbps mode that I believe is restricted to one port at a time. -- Blaine Gardner @ Evans & Sutherland 580 Arapeen Drive, SLC, Utah 84108 blgardne@javelin.sim.es.com BIX: blaine_g DoD #46 My other motorcycle is a Quadracer. FJ1200 Now I know why they are called BUTTERflys!