Skip to contentSkip to conversation

IDE is still crap.. or something

DISCUSSION Martin Eriksson 4 participants 9 messages Nov 23, 2001 – Nov 24, 2001

lore
Share & tools
lore
#0 lore raw

Well, just wanted to tell you that 2.4.15 still slows down to a crawl when copying a 500MB file between two hard drives (running ext3). I have tried any of the -c -u -m -W settings in hdparm. I even applied the 2.4.14 IDE patch (after fixing the rejects) but no go.

Even iptables is affected, because it takes forever to surf the internet from my behind-linux-firewall windows computer.

I'm right now trying to apply the preemptive-kernel patch to 2.4.15 but it had some strange rejects so it will be exciting to see if it works. I get good response from the -ac kernel series though.

_____________________________________________________

|  Martin Eriksson <[email protected]>
|  MSc CSE student, department of Computing Science
|  Umeå University, Sweden
#1 lore raw

----- Original Message -----
From: "Martin Eriksson" <[email protected]>
To: <[email protected]>
Sent: Friday, November 23, 2001 8:11 PM
Subject: IDE is still crap.. or something

Quoted 11 lines

Well, just wanted to tell you that 2.4.15 still slows down to a crawl when copying a 500MB file between two hard drives (running ext3). I have tried any of the -c -u -m -W settings in hdparm. I even applied the 2.4.14 IDE patch (after fixing the rejects) but no go.

Even iptables is affected, because it takes forever to surf the internet from my behind-linux-firewall windows computer.

I'm right now trying to apply the preemptive-kernel patch to 2.4.15 but it had some strange rejects so it will be exciting to see if it works. I get good response from the -ac kernel series though.

I applied the ide patch and the preemptive-kernel patch, and so far so good. Response is up again, but I'm not sure how well I fixed the .rej files. There were some reference to a "still_running" label that I simply ignored... *shrug*

Btw, i run my (pretty slow) hard disks on my BP6 HPT366 controller.

_____________________________________________________

|  Martin Eriksson <[email protected]>
|  MSc CSE student, department of Computing Science
|  Umeå University, Sweden
#2 lore raw

Just as an FYI...I am not seeing this on my ht366...runs nice and speedy...

Quoted 38 lines · Martin Eriksson

----- Original Message -----
From: "Martin Eriksson" <[email protected]>
To: <[email protected]>
Sent: Friday, November 23, 2001 8:11 PM
Subject: IDE is still crap.. or something

Quoted 11 lines

Well, just wanted to tell you that 2.4.15 still slows down to a crawl when copying a 500MB file between two hard drives (running ext3). I have tried any of the -c -u -m -W settings in hdparm. I even applied the 2.4.14 IDE patch (after fixing the rejects) but no go.

Even iptables is affected, because it takes forever to surf the internet from my behind-linux-firewall windows computer.

I'm right now trying to apply the preemptive-kernel patch to 2.4.15 but it had some strange rejects so it will be exciting to see if it works. I get good response from the -ac kernel series though.

I applied the ide patch and the preemptive-kernel patch, and so far so good. Response is up again, but I'm not sure how well I fixed the .rej files. There were some reference to a "still_running" label that I simply ignored... *shrug*

Btw, i run my (pretty slow) hard disks on my BP6 HPT366 controller.

_____________________________________________________

|  Martin Eriksson <[email protected]>
|  MSc CSE student, department of Computing Science
|  Ume� University, Sweden

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#3 lore raw
In reply to Martin Eriksson

Any previous kernel gave you good performance ?

Quoted 24 lines · Martin Eriksson

Well, just wanted to tell you that 2.4.15 still slows down to a crawl when copying a 500MB file between two hard drives (running ext3). I have tried any of the -c -u -m -W settings in hdparm. I even applied the 2.4.14 IDE patch (after fixing the rejects) but no go.

Even iptables is affected, because it takes forever to surf the internet from my behind-linux-firewall windows computer.

I'm right now trying to apply the preemptive-kernel patch to 2.4.15 but it had some strange rejects so it will be exciting to see if it works. I get good response from the -ac kernel series though.

_____________________________________________________

|  Martin Eriksson <[email protected]>
|  MSc CSE student, department of Computing Science
|  Umeå University, Sweden

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#4 lore raw

----- Original Message -----
From: "Marcelo Tosatti" <[email protected]>
To: "Martin Eriksson" <[email protected]>
Cc: "lkml" <[email protected]>
Sent: Friday, November 23, 2001 7:27 PM
Subject: Re: IDE is still crap.. or something

Quoted 2 lines

Any previous kernel gave you good performance ?

No.. not really. Except -ac kernels and kernels with the preempt patch.

I'll just clarify: The problem is not that hard disk operations gets slow (I can understand *that*, as it's a MW-DMA2 disk and one UDMA33 running ext3), instead the real problem is that the system as a *whole* gets painfully slow. Whilst copying a 500MB file, it takes ~6-8 seconds to start a new ssh session from my windows comp to my linux server, compared to <1 second when running preempt/ac kernels.

_____________________________________________________

|  Martin Eriksson <[email protected]>
|  MSc CSE student, department of Computing Science
|  Umeå University, Sweden
Quoted 4 lines · Martin Eriksson
Quoted 2 lines

Well, just wanted to tell you that 2.4.15 still slows down to a crawl

when

Quoted 1 line
Quoted 1 line

copying a 500MB file between two hard drives (running ext3). I have

tried

Quoted 7 lines
Quoted 7 lines

any of the -c -u -m -W settings in hdparm. I even applied the 2.4.14 IDE patch (after fixing the rejects) but no go.

Even iptables is affected, because it takes forever to surf the internet from my behind-linux-firewall windows computer.

I'm right now trying to apply the preemptive-kernel patch to 2.4.15 but

it

Quoted 1 line
Quoted 1 line

had some strange rejects so it will be exciting to see if it works. I

get

Quoted 1 line
Quoted 1 line

good response from the -ac kernel series though.

#5 lore raw
Quoted 2 lines · Martin Eriksson
Quoted 2 lines
Quoted 2 lines

any of the -c -u -m -W settings in hdparm. I even applied the 2.4.14 IDE patch (after fixing the rejects) but no go.

Mr. Martin Eriksson,

As for your subject -- "IDE" died a long time ago, but since it died before you entered university, I am not at all surprized. Now as for jumping on the case of the talented Mr. Marcelo Tosatti. He has not found it neccessary to enter university at this time, as he could likely teach the content scheduled in the next year to you.

Why are you doing thoughtless things like overriding the ruleset for optimizing the HOST/Device pair? I seriously doubt that you know the history of those option? Of the lot, one of them is retired as of ATA-2; however it still is optional for a while. The other is foolish in most cases unless dealing with ATA-2 hardware, or have audio driver problems. The next is settable by the kernel if you allow it to do the work for you. The last is also set by the kernel, should you allow it to operate. There is no valid reason for you to do anything w/ hdparm.

Now this is a global reply to your list of rants. Now if you can not merge patches and understand what is going on, then please keep the noise down. Next time please have some credablity when you attempt to make grand pontifications of code quality in Linux. Lastly you were not to be a target for everyones entertainment but this is where you have come.

Regards,

Andre Hedrick
Linux ATA Development
Linux Disk Certification Project

#6 lore raw

----- Original Message -----
From: "Andre Hedrick" <[email protected]>
To: "Martin Eriksson" <[email protected]>
Cc: "Marcelo Tosatti" <[email protected]>; "lkml" <[email protected]>
Sent: Saturday, November 24, 2001 8:06 AM
Subject: Re: IDE is still crap.. or something

Quoted 3 lines · Martin Eriksson
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

any of the -c -u -m -W settings in hdparm. I even applied the 2.4.14

IDE

Quoted 10 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line

patch (after fixing the rejects) but no go.

Mr. Martin Eriksson,

As for your subject -- "IDE" died a long time ago, but since it died before you entered university, I am not at all surprized. Now as for jumping on the case of the talented Mr. Marcelo Tosatti. He has not found it neccessary to enter university at this time, as he could likely teach the content scheduled in the next year to you.

Quoted 8 lines

Why are you doing thoughtless things like overriding the ruleset for optimizing the HOST/Device pair? I seriously doubt that you know the history of those option? Of the lot, one of them is retired as of ATA-2; however it still is optional for a while. The other is foolish in most cases unless dealing with ATA-2 hardware, or have audio driver problems. The next is settable by the kernel if you allow it to do the work for you. The last is also set by the kernel, should you allow it to operate. There is no valid reason for you to do anything w/ hdparm.

Well, actually I'm not the "must-use-hdparm -c1 -u1 -d1 -m16 -W1 -X66" kind of guy... I just tested some options because my system was slow. I do not run hdparm now, and everything works fine (with your ATA patch, and the preempt patch). I'm moving the hard disks to the "on-board" controller (PIIX) today, to see if that works better (without preempt+ata).

Also, I *would* be upgrading my linux system *if I had money*, but until then, I happily run with my crappy BP6, crappy HD's and crappy HPT366 controller.

Quoted 5 lines

Now this is a global reply to your list of rants. Now if you can not merge patches and understand what is going on, then please keep the noise down. Next time please have some credablity when you attempt to make grand pontifications of code quality in Linux. Lastly you were not to be a target for everyones entertainment but this is where you have come.

I'm sorry if the subject set you off.. what I should have written is propably "ATA hard disk access slows down my system", but I was tired and had previously been reading "comp.sys.ibm.pc.games.space-sim".

Also, I'm not meaning to sound important or anything with my sig. Maybe I should change it? Hmm...

Quoted 6 lines

Regards,

Andre Hedrick
Linux ATA Development
Linux Disk Certification Project

_____________________________________________________

|  Martin Eriksson <[email protected]>
|  Linux developer / Ignorant excuse for a human /
|  Ranting bastard / Swede
#7 lore raw

ATA is not crap. Propably.

In reply to Andre Hedrick

Sorry everyone I've upset.

The problem *seems* to be with my hard disks. Apparently both advertises udma2 capability but neither is capable?? When I run the disks on my "on-board" controller, I get a DMA timeout and the system comes to a halt (well only the HD activity, but you can't do much without access to the root/usr partition). When running on my HPT366 controller, I don't get the timeouts, but I do get the "slow responding system" at about the same time at which I get the timeouts on the PIIX4 controller. Problem comes when having copied about 11% of a 500'000'000 byte file from /dev/hdc7 to /dev/hda6...

The hard disks in question are models
hda: ST36451A
hdc: Maxtor 91152D8

The reason why I moved to the HPT366 controller in the first place, was because the onboard controller / BIOS? messed up the C/H/S values on one of the hard disks. Cannot be 100% sure though as this was a while ago (two years?).

_____________________________________________________

|  Martin Eriksson <[email protected]>
|  MSc CSE student, department of Computing Science
|  Umeå University, Sweden
#8 lore raw

Re: ATA is not crap. Propably.

Martin,

PIIX4 AB/EB is a broken hardware mess.
HPT366 is another issue.
The newest driver should improve things but the latest unreleased fixes a basic design error introduced some time ago by an unknown patch. This has been found and fixed. Just need to find an ungreased turkey first to test.

Regards,

Andre Hedrick
Linux ATA Development

Quoted 27 lines · Martin Eriksson

Sorry everyone I've upset.

The problem *seems* to be with my hard disks. Apparently both advertises udma2 capability but neither is capable?? When I run the disks on my "on-board" controller, I get a DMA timeout and the system comes to a halt (well only the HD activity, but you can't do much without access to the root/usr partition). When running on my HPT366 controller, I don't get the timeouts, but I do get the "slow responding system" at about the same time at which I get the timeouts on the PIIX4 controller. Problem comes when having copied about 11% of a 500'000'000 byte file from /dev/hdc7 to /dev/hda6...

The hard disks in question are models
hda: ST36451A
hdc: Maxtor 91152D8

The reason why I moved to the HPT366 controller in the first place, was because the onboard controller / BIOS? messed up the C/H/S values on one of the hard disks. Cannot be 100% sure though as this was a while ago (two years?).

_____________________________________________________

|  Martin Eriksson <[email protected]>
|  MSc CSE student, department of Computing Science
|  Ume� University, Sweden