Q: kallsyms - where can I find it and what does it do ?
Hello, I'm trying to build a new kernel with kdb support and I keep getting an error from the make file: kallsyms pass 1 [make] /bin/sh: /sbin/kallsyms: No such file or directory error What is kallsyms and where can I get it from ? Here is what I have and what I did: The machine is a Compaq Ap500 dual P-III Xeon. I'm using RedHat 6.2 (clean disk - custom install - install everything). The new kernel is linux-2.4.0-test9.tar.gz (the latest according to www.kernel.org). The kdb patch is kdb-v1.5-2.4.0-test9-pre9.gz (the latest according to oss.sgi.com). There doesn't seem to be a problem running the patch (no error messages, at least). I made sure kdb support is checked under 'make menuconfig'. I ran 'make dep; make clean; make bzImage' and it keeps failing. While on the subject, how is it possible that I can build an SMP kernel on that machine (before the patch), but can't build a UP kernel ?. I use 'make menuconfig' and uncheck "SMP support", then run 'make dep; make clean; make bzImage' and I get all kinds of warnings and errors until the make file simply stops running (doesn't return to the prompt with an error message - just stops and I have to hit CR to get back to the prompt). Thanks in advance, Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Please ignore - RE: kallsyms - where can I find it and what does it do ?
-Original Message- From: Hen, Shmulik [mailto:[EMAIL PROTECTED]] Sent: Wednesday, October 18, 2000 6:17 PM To: [EMAIL PROTECTED] Cc: 'Keith Owens' Subject: Q: kallsyms - where can I find it and what does it do ? Hello, I'm trying to build a new kernel with kdb support and I keep getting an error from the make file: kallsyms pass 1 [make] /bin/sh: /sbin/kallsyms: No such file or directory error What is kallsyms and where can I get it from ? Here is what I have and what I did: The machine is a Compaq Ap500 dual P-III Xeon. I'm using RedHat 6.2 (clean disk - custom install - install everything). The new kernel is linux-2.4.0-test9.tar.gz (the latest according to www.kernel.org). The kdb patch is kdb-v1.5-2.4.0-test9-pre9.gz (the latest according to oss.sgi.com). There doesn't seem to be a problem running the patch (no error messages, at least). I made sure kdb support is checked under 'make menuconfig'. I ran 'make dep; make clean; make bzImage' and it keeps failing. While on the subject, how is it possible that I can build an SMP kernel on that machine (before the patch), but can't build a UP kernel ?. I use 'make menuconfig' and uncheck "SMP support", then run 'make dep; make clean; make bzImage' and I get all kinds of warnings and errors until the make file simply stops running (doesn't return to the prompt with an error message - just stops and I have to hit CR to get back to the prompt). Thanks in advance, Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/ - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
page fault problems porting a network driver to 2.4.x
Hello,
We are developing an advanced networking services loadable module and are
having problems porting it to work on 2.4.x kernels. The driver is supposed
to provide services such as fault tolerance, load balancing and link
aggregation over a team of network adapters. It works OK on 2.2.x kernels
but hangs on 2.4.x kernels.
In order to debug it, we stripped it down to become a mere "intermediate" or
"filter" driver that binds to a base driver and passes everything through in
both directions (Rx, Tx, IOCTL, stats, etc.). After going through the basics
of modifying the driver to compile on 2.4.x kernels and fighting some nasty
dead locks due to the new nature of the networking layer, we managed to get
it to run. The driver will receive and transmit a few hundreds of thousands
of packets (while having a periodic timer expire 10 times a second and
running continuous IOCTLs), and then it causes an oops about not being able
to handle a page fault.
The function looks something like:
int iansHardStartXmit(struct sk_buff *skb, struct net_device *dev) {
int res;
struct net_device *base;
spin_lock(lock);
base = get_base_driver_by_name(name);
if(base != NULL) {
res = base-hard_start_xmit(skb, base);
}
spin_unlock(lock);
return res;
}
We used kdb in order to track down the problem and found out the following
stack trace:
EBPEIP function(args)
0xc4cd1c54 0xd081e3e7 [e100]__kallsyms+0xb (0xc4b595a0,
0xc840f200)
e100 __kallsyms 0xd081e3dc
0xd081e3dc 0xd0820dsc
0xd08244ba [ians]iansHardStartXmit+0xa6 (0xc4b595a0,
0xc4d9bc00)
ians .text 0xd0824060 0xd0824414
0xd082452c
0xc01f9d1f qdisc_restart+0xcf (0xc4d9bc00)
kernel .text 0xc010 0xc01f9c50
0xc01f9f14
*
*
*
This goes on and shows that this is an ICMP echo reply packet going down
through the IP stack to the filter driver (apparently 0xc4b595a0 is the skb,
0xc4d9bc00 is the *dev of the filter driver and 0xc840f200 is the *dev of
the base driver). The filter driver is supposed to call the
dev-hard_start_xmit of the base driver, but strangely it lands somewhere in
the data segment of the base driver (__kallsyms is a part of the symbol
table of the module according to insmod -m).
Figuring the dev-hard_start_xmit pointer got trashed somehow, we added a
check to make sure the same pointer is always called, and indeed this was
the case. Looking at the assembly code with kdb, we could see that the call
to the base driver is done by a 'call *%eax' command. kdb reports that
eax=0x after the page fault (origeax).
How is it possible that the pointer to the function keeps it's value, but
the jump to that function falls somewhere else ?
The entire function is protected by a spinlock, so there is no worry about
the other threads messing my data.
We are using:
RedHat 6.2
gcc v2.91.66
modutils v2.3.11-1
kernel linux-2.4.0-test9
kdb v1.5-2.4.0-test9-pre9
Compaq ap500 dual p-III Xeon
Thanks,
Shmulik Hen
Software Engineer
Linux Advanced Networking Services
Network Communications Group, Israel (NCGj)
Intel Corporation Ltd.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
Please read the FAQ at http://www.tux.org/lkml/
pointer to dev-hard_start_xmit() gets trashed in 2.4.0-test9
Hello,
We are developing an advanced networking services loadable module and are
having problems porting it to work on 2.4.x kernels. The driver is supposed
to provide services such as fault tolerance, load balancing and link
aggregation over a team of network adapters. It works OK on 2.2.x kernels
but hangs on 2.4.x kernels.
In order to debug it, we stripped it down to become a mere "intermediate" or
"filter" driver that binds to a base driver and passes everything through in
both directions (Rx, Tx, IOCTL, stats, etc.). After going through the basics
of modifying the driver to compile on 2.4.x kernels and fighting some nasty
dead locks due to the new nature of the networking layer, we managed to get
it to run. The driver will receive and transmit a few hundreds of thousands
of packets (while having a periodic timer expire 10 times a second and
running continuous IOCTLs), and then it causes an oops about not being able
to handle a page fault.
The function looks something like:
int iansHardStartXmit(struct sk_buff *skb, struct net_device *dev) {
int res;
struct net_device *base;
spin_lock(lock); //no interrupts involved, so spin_lock
should do
base = get_base_driver_by_name(name);
if(base != NULL) {
BUG_TRAP(ptr_g == base-hard_start_xmit); //make sure it's
always the same addr
res = base-hard_start_xmit(skb, base);
}
spin_unlock(lock);
return res;
}
We used kdb in order to track down the problem and found out the following
stack trace:
EBPEIP function(args)
0xc4cd1c54 0xd081e3e7 [e100]__kallsyms+0xb (0xc4b595a0,
0xc840f200)
e100 __kallsyms 0xd081e3dc
0xd081e3dc 0xd0820dsc
0xd08244ba [ians]iansHardStartXmit+0xa6 (0xc4b595a0,
0xc4d9bc00)
ians .text 0xd0824060 0xd0824414
0xd082452c
0xc01f9d1f qdisc_restart+0xcf (0xc4d9bc00)
kernel .text 0xc010 0xc01f9c50
0xc01f9f14
*
*
*
This goes on and shows that this is an ICMP echo reply packet going down
through the IP stack to the filter driver (apparently 0xc4b595a0 is the skb,
0xc4d9bc00 is the *dev of the filter driver and 0xc840f200 is the *dev of
the base driver). The filter driver is supposed to call the
dev-hard_start_xmit of the base driver, but strangely it lands somewhere in
the data segment of the base driver (__kallsyms is a part of the symbol
table of the module according to insmod -m).
Figuring the dev-hard_start_xmit pointer got trashed somehow, we added a
check to make sure the same pointer is always called, and indeed this is the
case. Looking at the assembly code with kdb, we could see that the call to
the base driver is done by a 'call *%eax' instruction.
How is it possible that the pointer to the function keeps it's value, but
the jump to that function falls somewhere else ?
We are using:
RedHat 6.2
gcc v2.91.66
modutils v2.3.11-1 (was upgraded because of kdb)
kernel linux-2.4.0-test8 (SMP, +kdb, compiled with frame pointers,
SPINLOCK_DEBUG=2)
kdb v1.4-2.4.0-test9-pre9
Compaq ap500 dual p-III Xeon
Could this be a version mismatch between the components above ?
Thanks,
Shmulik Hen
Software Engineer
Linux Advanced Networking Services
Network Communications Group, Israel (NCGj)
Intel Corporation Ltd.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
Please read the FAQ at http://www.tux.org/lkml/
RE: spinlock help
How about if the same sequence occurred, but from two different drivers ?
We've had some bad experience with this stuff. Our driver, which acts as an
intermediate net driver, would call the hard_start_xmit in the base driver.
The base driver, wanting to block receive interrupts would issue a
'spin_lock_irqsave(a,b)' and process the packet. If the TX queue is full, it
could call an indication entry point in our intermediate driver to signal it
to stop sending more packets. Since our indication function handles many
types of indications but can process them only one at a time, we wanted to
block other indications while queuing the request.
The whole sequence would look like that:
[our driver]
ans_send() {
.
.
e100_hard_start_xmit(dev, skb);
.
.
}
[e100.o]
e100_hard_start_xmit() {
.
.
spin_lock_irqsave(a,b);
.
.
if(tx_queue_full)
ans_notify(TX_QUEUE_FULL); --
.
.
spin_unlock_irqrestore(a,b);
}
[our driver]
ans_notify() {
.
.
spin_lock_irqsave(c,d);
queue_request(req_type);
spin_unlock_irqrestore(c,d);--
.
.
}
At that point, for some reason, interrupts were back and the e100.o would
hang in an infinite loop (we verified it on kernel 2.4.0-test10 +kdb that
the processor was enabling interrupts and that the e100_isr was called for
processing an Rx int.).
How is that possible that a 'spin_unlock_irqrestore(c,d)' would also restore
what should have been restored only with a 'spin_unlock_irqrestore(a,b)' ?
Thanks in advance,
Shmulik Hen
Software Engineer
Linux Advanced Networking Services
Intel Network Communications Group
Jerusalem, Israel.
-Original Message-
From: Nigel Gamble [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 1:54 AM
To: Manoj Sontakke
Cc: [EMAIL PROTECTED]
Subject: Re: spinlock help
On Tue, 6 Mar 2001, Manoj Sontakke wrote:
1. when spin_lock_irqsave() function is called the subsequent code is
executed untill spin_unloc_irqrestore()is called. is this right?
Yes. The protected code will not be interrupted, or simultaneously
executed by another CPU.
2. is this sequence valid?
spin_lock_irqsave(a,b);
spin_lock_irqsave(c,d);
Yes, as long as it is followed by:
spin_unlock_irqrestore(c, d);
spin_unlock_irqrestore(a, b);
Nigel Gamble[EMAIL PROTECTED]
Mountain View, CA, USA. http://www.nrg.org/
-
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/
-
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/
RE: spinlock help
spin_lock_bh() won't block interrupts and we need them blocked to prevent
more indications.
spin_lock_irq() could do the trick but it's counterpart spin_unlock_irq()
enables all interrupts by calling sti(), and this is even worse for us.
Shmulik.
-Original Message-
From: Manoj Sontakke [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 12:27 PM
To: Hen, Shmulik
Cc: '[EMAIL PROTECTED]'; Manoj Sontakke; [EMAIL PROTECTED]
Subject: Re: spinlock help
hi
spin_lock_irq() andspin_lock_bh()
can they be of any use to u?
"Hen, Shmulik" wrote:
How about if the same sequence occurred, but from two different drivers ?
We've had some bad experience with this stuff. Our driver, which acts as
an
intermediate net driver, would call the hard_start_xmit in the base
driver.
The base driver, wanting to block receive interrupts would issue a
'spin_lock_irqsave(a,b)' and process the packet. If the TX queue is full,
it
could call an indication entry point in our intermediate driver to signal
it
to stop sending more packets. Since our indication function handles many
types of indications but can process them only one at a time, we wanted to
block other indications while queuing the request.
The whole sequence would look like that:
[our driver]
ans_send() {
.
.
e100_hard_start_xmit(dev, skb);
.
.
}
[e100.o]
e100_hard_start_xmit() {
.
.
spin_lock_irqsave(a,b);
.
.
if(tx_queue_full)
ans_notify(TX_QUEUE_FULL); --
.
.
spin_unlock_irqrestore(a,b);
}
[our driver]
ans_notify() {
.
.
spin_lock_irqsave(c,d);
queue_request(req_type);
spin_unlock_irqrestore(c,d);--
.
.
}
At that point, for some reason, interrupts were back and the e100.o would
hang in an infinite loop (we verified it on kernel 2.4.0-test10 +kdb that
the processor was enabling interrupts and that the e100_isr was called for
processing an Rx int.).
How is that possible that a 'spin_unlock_irqrestore(c,d)' would also
restore
what should have been restored only with a 'spin_unlock_irqrestore(a,b)' ?
Thanks in advance,
Shmulik Hen
Software Engineer
Linux Advanced Networking Services
Intel Network Communications Group
Jerusalem, Israel.
-Original Message-
From: Nigel Gamble [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 1:54 AM
To: Manoj Sontakke
Cc: [EMAIL PROTECTED]
Subject: Re: spinlock help
On Tue, 6 Mar 2001, Manoj Sontakke wrote:
1. when spin_lock_irqsave() function is called the subsequent code is
executed untill spin_unloc_irqrestore()is called. is this right?
Yes. The protected code will not be interrupted, or simultaneously
executed by another CPU.
2. is this sequence valid?
spin_lock_irqsave(a,b);
spin_lock_irqsave(c,d);
Yes, as long as it is followed by:
spin_unlock_irqrestore(c, d);
spin_unlock_irqrestore(a, b);
Nigel Gamble[EMAIL PROTECTED]
Mountain View, CA, USA. http://www.nrg.org/
-
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/
-
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/
--
Regards,
Manoj Sontakke
-
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/
RE: spinlock help
e100 implements all sorts of hooks for our intermediate driver (kind of a
co-development effort), so eepro100 is out of the question for us.
Shmulik.
-Original Message-
From: Ofer Fryman [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 12:31 PM
To: 'Hen, Shmulik'
Cc: [EMAIL PROTECTED]
Subject: RE: spinlock help
Did you try looking at Becker eepro100 driver it seems to be simple, no
unnecessary spin_lock_irqsave?.
-Original Message-
From: Hen, Shmulik [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 11:21 AM
To: '[EMAIL PROTECTED]'; Manoj Sontakke
Cc: [EMAIL PROTECTED]
Subject: RE: spinlock help
How about if the same sequence occurred, but from two different drivers ?
We've had some bad experience with this stuff. Our driver, which acts as an
intermediate net driver, would call the hard_start_xmit in the base driver.
The base driver, wanting to block receive interrupts would issue a
'spin_lock_irqsave(a,b)' and process the packet. If the TX queue is full, it
could call an indication entry point in our intermediate driver to signal it
to stop sending more packets. Since our indication function handles many
types of indications but can process them only one at a time, we wanted to
block other indications while queuing the request.
The whole sequence would look like that:
[our driver]
ans_send() {
.
.
e100_hard_start_xmit(dev, skb);
.
.
}
[e100.o]
e100_hard_start_xmit() {
.
.
spin_lock_irqsave(a,b);
.
.
if(tx_queue_full)
ans_notify(TX_QUEUE_FULL); --
.
.
spin_unlock_irqrestore(a,b);
}
[our driver]
ans_notify() {
.
.
spin_lock_irqsave(c,d);
queue_request(req_type);
spin_unlock_irqrestore(c,d);--
.
.
}
At that point, for some reason, interrupts were back and the e100.o would
hang in an infinite loop (we verified it on kernel 2.4.0-test10 +kdb that
the processor was enabling interrupts and that the e100_isr was called for
processing an Rx int.).
How is that possible that a 'spin_unlock_irqrestore(c,d)' would also restore
what should have been restored only with a 'spin_unlock_irqrestore(a,b)' ?
Thanks in advance,
Shmulik Hen
Software Engineer
Linux Advanced Networking Services
Intel Network Communications Group
Jerusalem, Israel.
-Original Message-
From: Nigel Gamble [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 1:54 AM
To: Manoj Sontakke
Cc: [EMAIL PROTECTED]
Subject: Re: spinlock help
On Tue, 6 Mar 2001, Manoj Sontakke wrote:
1. when spin_lock_irqsave() function is called the subsequent code is
executed untill spin_unloc_irqrestore()is called. is this right?
Yes. The protected code will not be interrupted, or simultaneously
executed by another CPU.
2. is this sequence valid?
spin_lock_irqsave(a,b);
spin_lock_irqsave(c,d);
Yes, as long as it is followed by:
spin_unlock_irqrestore(c, d);
spin_unlock_irqrestore(a, b);
Nigel Gamble[EMAIL PROTECTED]
Mountain View, CA, USA. http://www.nrg.org/
-
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/
-
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/
-
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/
RE: spinlock help
The kdb trace was accurate, we could actually see the e100 ISR pop from no
where right in the middle of our ans_notify every time the TX queue would
fill up. When we commented out the call to spin_*_irqsave(), it worked fine
under heavy stress for days.
Is it possible it was something wrong with 2.4.0-test10 specifically ?
We had to drop the locks in the final release and never got around to
checking it on other kernel releases (it went on the TO_DO list ;-).
Shmulik.
-Original Message-
From: Andrew Morton [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 1:43 PM
To: Hen, Shmulik
Subject: Re: spinlock help
"Hen, Shmulik" wrote:
How about if the same sequence occurred, but from two different drivers ?
We've had some bad experience with this stuff. Our driver, which acts as
an
intermediate net driver, would call the hard_start_xmit in the base
driver.
The base driver, wanting to block receive interrupts would issue a
'spin_lock_irqsave(a,b)' and process the packet. If the TX queue is full,
it
could call an indication entry point in our intermediate driver to signal
it
to stop sending more packets. Since our indication function handles many
types of indications but can process them only one at a time, we wanted to
block other indications while queuing the request.
The whole sequence would look like that:
[our driver]
ans_send() {
.
.
e100_hard_start_xmit(dev, skb);
.
.
}
[e100.o]
e100_hard_start_xmit() {
.
.
spin_lock_irqsave(a,b);
.
.
if(tx_queue_full)
ans_notify(TX_QUEUE_FULL); --
.
.
spin_unlock_irqrestore(a,b);
}
[our driver]
ans_notify() {
.
.
spin_lock_irqsave(c,d);
queue_request(req_type);
spin_unlock_irqrestore(c,d);--
.
.
}
At that point, for some reason, interrupts were back
Sorry, that can't happen.
Really, you must have made a mistake somewhere.
-
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/
RE: spinlock help
OK guys, you were right. The bug was in our code - sorry for trouble. Turns out that while I was away, the problem was solved by someone else. The problem is probably related to the fact that when we did 'spin_lock_irqsave(c,d)', 'd' was a global variable. The fix was to wrap the call with another function and declare 'd' as local. I can't quite explain, but I think that changing from a static to automatic variable made the difference. My best guess is that since 'd' is passed by value and not by reference, the macro expansion of spin_lock_irqsave() relies on the location of 'd' in the stack and if 'd' was on the heap instead, it might get trashed. I would really like to hear your expert opinion on my assumption. Thanks, Shmulik. -Original Message- From: Andrew Morton [mailto:[EMAIL PROTECTED]] Sent: Wednesday, March 07, 2001 2:54 PM To: Hen, Shmulik Cc: 'LKML' Subject: Re: spinlock help "Hen, Shmulik" wrote: The kdb trace was accurate, we could actually see the e100 ISR pop from no where right in the middle of our ans_notify every time the TX queue would fill up. When we commented out the call to spin_*_irqsave(), it worked fine under heavy stress for days. Is it possible it was something wrong with 2.4.0-test10 specifically ? Sorry, no. If spin_lock_irqsave()/spin_unlock_irqrestore() were accidentally reenabling interrupts then it would be the biggest, ugliest catastrophe since someone put out a kernel which forgot to flush dirty inodes to disk :) Conceivably it was a compiler bug. Were you using egcs-1.1.2/x86? - - 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/
catch 22 - porting net driver from 2.2 to 2.4
Hello, This is a bit long and I apologize (since there are kdb captures in it). We are developing an advanced networking services driver (loadable module) and are having problems porting it to work on 2.4.x kernel. The driver is supposed to provide services such as fault tolerance, load balancing, link aggregation and VLAN. It does that by creating a group of "virtual" adapters that are bound on top of a team of network adapters. This works great on 2.2 kernels but demonstrated a few problems on 2.4.0-test9 The only problem we have left now has to do with insmod/rmmod. for good reasons, we cant just call init_etherdev() like base drivers do, so we created our own version that handles memory and name allocations and calls register_netdevice() on it's own the same as init_etherdev(). since we've got several "virtual" adapters that are part of a topology being built progressively, we can't perform their registrations during module_init() but rather through an IOCTL and here is our problem: if we call register_netdevice(), we get the following message: RTNL: assertion failed at devinet.c(775):inetdev_event we figured this is because we neglected rtnl_lock() so instead we try using register_netdev() to handle this for us but then we get: Scheduling in interrupt kernel BUG at sched.c:696! Entering kdb (current=0xc51a8000, pid 1075) on processor 1 Panic: invalid operand due to panic @ 0xc011aa71 eax = 0x001b ebx = 0x0020 ecx = 0xc030d80c edx = 0x esi = 0xc030b5b4 edi = 0x esp = 0xc51a9d34 eip = 0xc011aa71 ebp = 0xc51a9d8c ss = 0x0018 cs = 0x0010 eflags = 0x00010246 ds = 0x0018 es = 0x0018 origeax = 0x regs = 0xc51a9d00 [1]kdb bt EBP EIP Function(args) 0xc51a9d8c 0xc011aa71 schedule+0x935 (0xc2c9e000, 0xc4482520, 0xc51a9e10) kernel .text 0xc010 0xc011a13c 0xc011aa80 0xc51a9db8 0xc0107b8d __down+0xf5 kernel .text 0xc010 0xc0107a98 0xc0107c68 0xc51a9dcc 0xc0107f43 __down_failed+0xb (0xc51a9de4, 0xd082de5c, 0xc2c9e000, 0xc60a8320, 0xc51a9dfc) kernel .text 0xc010 0xc0107f38 0xc0107f4c 0xc023a7a9 stext_lock+0x4919 kernel .text.lock 0xc0235e90 0xc0235e90 0xc023bd80 0xc51a9dd4 0xc01f2e81 rtnl_lock+0x11 (0xc2c9e000) kernel .text 0xc010 0xc01f2e70 0xc01f2e88 0xc51a9de4 0xd082de5c [ians]iansInitEtherdev+0x20 (0xc4482520) ians .text 0xd082d060 0xd082de3c 0xd082de78 . . (boring chain of calls) . 0xc51a9ec4 0xd082dbcd [ians]doControlIoctl+0x15d (0xc2c9e200, 0xc51a9f20, 0x89f0) ians .text 0xd082d060 0xd082da70 0xd082dc40 0xc51a9ee4 0xc01ef09f dev_ifsioc+0x33f (0xc51a9f20, 0x89f0, 0xc51a9f20) kernel .text 0xc010 0xc01eed60 0xc01ef0b0 0xc51a9f40 0xc01ef29d dev_ioctl+0x1ed (0x89f0, 0xba58) kernel .text 0xc010 0xc01ef0b0 0xc01ef300 0xc51a9f64 0xc021a70c inet_ioctl+0x18c (0xc339d13c, 0x89f0, 0xba58) kernel .text 0xc010 0xc021a580 0xc021a720 0xc51a9f84 0xc01e8f06 sock_ioctl+0x9e (0xc339d040, 0xc38e0900, 0x89f0, 0xba58) kernel .text 0xc010 0xc01e8e68 0xc01e8f6c 0xc51a9fbc 0xc014f5fd sys_ioctl+0x26d (0x3, 0x89f0, 0xba58, 0x4000ae60, 0xbba4) kernel .text 0xc010 0xc014f390 0xc014f6a0 0xc010965f system_call+0x33 kernel .text 0xc010 0xc010962c 0xc0109664 We figured that since we are in user context (do_ioctl) and use spin_lock_bh() to protect us from other concurrent threads, it might interfere with rtnl_lock() so we remove our lock just before calling register_netdev() and lock again upon return but then the whole process just stopped and didn't return to the prompt. from within kdb, we could see that all CPU's are running in idle but if we try to return to the prompt the whole system hangs. sometimes it hangs if we try to run ifconfig -a to see if the virtual adapters appear. I can't use it without locks, I can't use it with locks and I can't complete the operation if I remove my own locks - catch 22. The other problem has to do with rmmod - here we get called in our cleanup_module function and from it we try to call unregister_netdev() for each registered virtual adapter. in this case, we get: Scheduling in interrupt kernel BUG at sched.c:696! Entering kdb (current=0xc38c8000, pid 1602) on processor 0 Panic:
Multiple warnings when compiling network driver in 2.4.0-test9
Hello,
While trying to compile a network driver for 2.4.0-test9 (+kdb-v1.5,
configured for UP) I'm getting multiple warnings:
/usr/src/linux/include/linux/sched.h:700: warning: can't inline call
to `__mmdrop'
/usr/src/linux/include/linux/sched.h:704: warning: called from here
This happens for every kernel header I try to use such as netdevice.h,
skbuff.h, malloc.h and pci.h
Each of those header files includes slab.h (line 14) that includes mm.h
(line 4) that includes sched.h which contains the following on line 700:
700:extern inline void FASTCALL(__mmdrop(struct mm_struct *));
701:static inline void mmdrop(struct mm_struct * mm)
702:{
703:if (atomic_dec_and_test(mm-mm_count))
704:__mmdrop(mm);
705:}
My make file uses the following flags:
gcc -fomit-frame-pointer -Wall -Wstrict-prototypes -Winline -O3
-D__KERNEL__ -DMODULE -DDEBUG -DMODVERSIONS -I/usr/src/linux/include
Can anyone tell me what this warning means and if I can safely ignore it (or
expect disaster) ?
Thanks,
Shmulik Hen,
Software Engineer
Linux Advanced Networking Services
Network Communications Group, Israel (NCGj)
Intel Corporation Ltd.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
Please read the FAQ at http://www.tux.org/lkml/
Locking Between User Context and Soft IRQs in 2.4.0
Hello, We are trying to port a network driver from 2.2.x to 2.4.x and have some question regarding locks. According to the kernel locking HOWTO, we have to take extra care when locking between user context threads and BH/tasklet/softIRQ, so we learned (the hard way ;-) that when running the ioctl system call from an application we should use spin_lock/unlock_bh() and not spin_lock/unlock() inside dev-do_ioctl(). * What about the other entry points implemented in net_device ? * We've got dev-get_stats, dev-set_mac_address, dev-set_mutlicast_list and others that are all called from running 'ifconfig' which is an application. Are they considered user context too ? * What about dev-open and dev-stop ? * We figured that dev-hard_start_xmit() and timer callbacks are not considered user context, but how can I find out if they are being run as SoftIRQ or as tasklets or as Bottom Halves ? (their different definitions require different types of protections) Our driver is actually an intermediate driver bound on top of a regular net driver. It behaves both as a network adapter driver and a protocol at the same time. I can safely assume that it will have to handle both transmits and receives simultaneously (no hardware interrupts are involved). We've decided that for the first stage we are going to implement "wide" locks that wrap entire operations from top to bottom. For example, our dev-hard_start_xmit() will have a spin_lock() at the beginning and a spin_unlock() at the end of the function. * Will it be safe to keep the lock until after the call to the base driver's hard_start_xmit, or do I have to release the lock just before that ? * Or, in our receive function, will I have to release the lock before or after the call to netif_rx() ? * What about other calls to the kernel ? can the running thread be switched out of context when calling kernel entries and not be switched back in when they finish ? should I beware of deadlocks in such case ? Thanks in advance, Shmulik Hen, Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: catch 22 - porting net driver from 2.2 to 2.4
and you don't get the "RTNL: assertion failed at devinet.c(775):inetdev_event" in 2.4.x ? the thing is I need to prevent Tx/Rx when a topology change is initiated from the ioctl (registering a virtual adapter is just one example), so they all share a single lock and I must use spin_lock_bh from the ioctl. Shmulik. -Original Message- From: Olaf Titz [mailto:[EMAIL PROTECTED]] Sent: Friday, November 10, 2000 2:09 AM To: [EMAIL PROTECTED] Subject: Re: catch 22 - porting net driver from 2.2 to 2.4 We figured that since we are in user context (do_ioctl) and use spin_lock_bh() to protect us from other concurrent threads, it might interfere with rtnl_lock() so we remove our lock just before calling register_netdev() and lock again upon return but then the whole process just stopped and didn't return to the prompt. from within kdb, we could see that Can't you just do this: #if LINUX_VERSION_CODE = KERNEL_VERSION(2,3,0) /* not sure about the 0 */ #define rtnl_LOCK() rtnl_lock() #define rtnl_UNLOCK() rtnl_unlock() #else #define rtnl_LOCK() /* nop */ #define rtnl_UNLOCK() /* nop */ #endif rtnl_LOCK(); register_netdevice(...); rtnl_UNLOCK(); that works for me (yes, from do_ioctl, but without the bh lock - I don't know if that's absolutely needed in your case). Olaf - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/ - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: catch 22 - porting net driver from 2.2 to 2.4
So how come I get the "RTNL: assertion failed at devinet.c(775):inetdev_event" when I call register_netdevice without rtnl_lock/unlock ? could it be a 2.4.0-test9 thing ? (haven't used test10 or 11 yet). and what about rmmod causing the panic when I use unregister_netdev or never completing the operation when I use unregister_netdevice ? does module_exit run inside rtnl_lock too ? Shmulik. -Original Message- From: Jeff Garzik [mailto:[EMAIL PROTECTED]] Sent: Thursday, November 09, 2000 7:37 PM To: Hen, Shmulik Cc: 'LNML'; 'LKML'; [EMAIL PROTECTED] Subject: Re: catch 22 - porting net driver from 2.2 to 2.4 do_ioctl is inside rtnl_lock... Remember if you need to alter the rules, you can always queue work in the current context, and have a kernel thread handle the work. The nice thing about a kernel thread is that you start with a [almost] clean state, when it comes to locks. Jeff -- Jeff Garzik | Building 1024 | Would you like a Twinkie? MandrakeSoft| - To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to [EMAIL PROTECTED] - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: catch 22 - porting net driver from 2.2 to 2.4
"Jeff Garzik" wrote: Theoretically, if you call unregister_netdev from rmmon, it should grab rtnl_lock and then complete the operation for you. If that doesn't work for you, it sounds like you are not setting up, or cleaning up, something correctly. Basically... it sounds like there are still bugs in your driver that need working out :) I followed the value of dev-refcnt and there is something strange. before the call to register_netdev it is set to 0 and after that it is increased to 1. but before the call to unregister_netdev it is somehow 2. How can I tell who is modifying it and when ? I tried using kdb to find out but it keeps hanging the machine. I wanted to place a breakpoint that will pop if a certain memory address is accessed so I did the following: static void *p; (global - at the top of my source file) EXPORT_SYMBOL (p); in my probe function: p = (void*) dev-refcnt; insmod my_module ksyms this showed that p is at address 0xd081ee90 then from within kdb: kdb md 0xd081ee90 0xd081ee90 c470bedc kdb bpha 0xc470bedc Forced Global Breakpoint at... kdb go system hung Is there a way to place a breakpoint on a memory address access ? - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Q: using kdb to trap memory modifications
hello, I'm trying to see when and how a certain variable is being modified and I wonder how to get kdb to do that for me. when I do 'insmod my_module' with my network driver, I noticed that dev-refcnt is 0 at first and gets increased to 1 after calling register_netdev. When I want to do 'rmmod my_module' somehow dev-refcnt is at 2 before the call to unregister_netdev and so the kernel (2.4.0-test9) won't let the module to unload. Since I don't increase it explicitly, I want to know who does it for me and when to see if I'm doing anything wrong. So, I modified my module to contain in the global section: static void* ptr; EXPORT_SYMBOL(ptr); and in my module_init function I added: ptr = (void*) dev-refcnt; after running 'insmod my_module' I used ksyms and found that ptr is at address 0xd081eeb4. after entering kdb, I did: kdb md 0xd081eeb4 0xd081eeb4 c470bedc 0xd081eec4 . . kdb bph 0xc470bedc Forced Breakpoint at #0 ... kdb be 0 kdb go When I try to return to normal shell, the system is totally hung and won't even receive inputs from the remote serial terminal. I tried bp, bpa and bpha, but the result is always the same. Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: catch 22 - porting net driver from 2.2 to 2.4
I tried using the kernel thread as demonstrated in your example and again it failed (panic - scheduling in interrupt). The difference is that your code executes the thread from within dev-open, while my code tries to do that from dev-do_ioctl that has spinlocks around the entire operation (which apparently sleeps). If I comment out the spin_lock/unlock it will succeed, but then I can't be sure I don't get any concurrent Tx/Rx/timer which is a bad idea while the topology is still being created. is there any way to do something like firing threads/timers atomically ? Thanks, Shmulik. -Original Message- From: Jeff Garzik [mailto:[EMAIL PROTECTED]] Sent: Monday, November 13, 2000 4:26 PM To: Hen, Shmulik Subject: Re: catch 22 - porting net driver from 2.2 to 2.4 "Hen, Shmulik" wrote: Where can I find info about that ? My first idea was to fire a timer and let the callback routine do the work, but I worry about synchronization and about passing the list of items for it to handle. What is the accepted way of starting a kernel thread and how do I handle parameters and sync. ? Attached is an example. My "8139too" ethernet driver uses a kernel thread instead of a timer to perform media checking. It illustrates how to start and stop a kernel thread. -- Jeff Garzik | Building 1024 | The chief enemy of creativity is "good" sense MandrakeSoft| -- Picasso - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: change_mtu boundary checking error
But Ethernet is not only for IP, what about other protocols ? -Original Message- From: Alan Cox [mailto:[EMAIL PROTECTED]] Sent: Tuesday, April 17, 2001 3:41 PM To: [EMAIL PROTECTED] Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: Re: change_mtu boundary checking error Now, the high boundary seemed reasonable (ETH_FRAME_LEN - ETH_HLEN = ETH_DATA_LEN) which gives 1500, but why is the low boundary set to 68 ? According to my calculations, it should have been ETH_ZLEN - ETH_HLEN which gives 46. The IPv4 minimum MTU is 68 bytes. Below that not all frames can be delivered - 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/
RE: ioctl call for network device
struct ifreq has a member called ifr_data. It is a pointer. You can put a pointer to any of your data, including the most complex structure you might envision, in that area. This allows you to pass anything to and from your module. This pointer can be properly dereferenced in kernel space but you should use copy_to/from_user and friends so a user-space coding bug won't panic the kernel. How about a linked list ? Will the driver be able to follow the list where each node was dynamically allocated by the application ? Is there a size limit on the buffer ifr_data points to ? (AFAIK, Windows NDIS drivers limit to 1 page buffer =4096 bytes). Thanks, Shmulik Hen Linux Advanced Networking Services Intel Network Communications Group - 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/
[OT] ethtool MII helpers (actually two OT's)
MII --- Is there any support in the MII standard for 1000Mbps (GbE Fiber/Copper) ? Perhaps an extension to the standard ? I could see that some of the Gigabit adapters supported by the kernel provide the MII IOCTLs interface, but couldn't figure out how to extract the correct speed information from the registers I can read. I know it's a bit of a hassle and I have to get the local capabilities and match them against the partner's capabilities and find the highest common speed etc. etc. but I'm sure that if the driver can do it I can reproduce it in userland too. EthTool --- Is there a way that I can extract the link status information out of the ethtool struct ? I could see that at least one Gigabit adapter driver (bcm5700.c), provides the EthTool interface and reports the correct speed and duplex mode but not the link status. Is there a place that defines how a driver is supposed to implement the support for EthTool ? I figured that since there is no separate field for link status (at least in version 1.2), a driver is supposed to report speed=0 or something like that when the link is down. I know this driver detects link status changes for sure because it prints messages every time, but the speed and duplex are always reported the same. Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Intel Network Communications Group Jerusalem, Israel -Original Message- From: Jeff Garzik [mailto:[EMAIL PROTECTED]] Sent: Friday, June 22, 2001 8:59 AM To: Chris Wedgwood Cc: Linux Kernel Mailing List; [EMAIL PROTECTED]; David S. Miller Subject: Re: PATCH: ethtool MII helpers Chris Wedgwood wrote: On Fri, Jun 22, 2001 at 01:24:36AM -0400, Jeff Garzik wrote: Sure, and that's planned. Wanna send me a patch for it? :) Possibly, but I wonder if this is a kernel-space problem or not. Why not put all the smarts into userland for it? I meant, send me a patch for userland ethtool, to do exactly what you described. It will definitely fall back on the MII ioctls if ethtool media support for the desired command doesn't exist. Well, that is more or less as much as needs to be done. That, and some kind of super-set API to be defined for all new stuff, having two slightly different APIs for the same things sucks. Both APIs do different things but have a common subset, yes. The MII ioctls only do their thing for MII-like hardware. ethtool can be applied to any hardware. Old ISA drivers that don't do MII, or do it in a really nonstandard way. For example I have ethtool code locally which allows ne2k-pci to do media selection via ioctl, for two popular ne2k cards, something its never been able to do before. Emulating media selection support for things like 10base2-10baseT-AUI just isn't possible with the MII ioctls. MII is a standard and incredibly popular, thus mii-tool works most popular PCI NICs, for the most popular media types. But it's still basically a hardware interface. I am not convinced its a good idea for make the [G]MII ioctls the Linux software media interface for all network hardware. I see ethtool as the interface for tuning your NIC, that works across all hardware. I see mii-diag as the way to do advance MII-specific hardware stuff, like next page or HA monitoring or whatever. Jeff -- Jeff Garzik | Andre the Giant has a posse. Building 1024| MandrakeSoft | - 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/ - 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/
kernel memory allocations alignment
Hello, When using kmalloc(size_t size), do I get a guaranty that the memory region allocated is aligned according to the size specified ? More to the point, if I call kmalloc for type int on an IA64 architecture is the pointer going to be 8 bytes aligned ? Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: kernel memory allocations alignment
Actually yes. We were warned that on IA64 architecture the system will halt when accessing any type of variable via a pointer if the pointer does not contain an aligned address matching that type. Until now we were using a method of receiving a pointer to an array, casting it to a pointer of a struct (packed with #pragma pack(1) ) ,and retrieving fields directly from it with pointers. It seems we cannot do that any more and were wondering what are the alternatives. One way we could think of is forget the packing and rearrange the fields in the struct in descending order so they all come out aligned, but we didn't know for sure if the first one will be aligned too. Will that work ? Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Intel Network Communications Group Jerusalem, Israel -Original Message- From: Manfred [mailto:[EMAIL PROTECTED]] Sent: Sunday, February 04, 2001 5:56 PM To: Hen, Shmulik Cc: 'LKML' Subject: Re: kernel memory allocations alignment "Hen, Shmulik" wrote: When using kmalloc(size_t size), do I get a guaranty that the memory region allocated is aligned according to the size specified ? More to the point, if I call kmalloc for type int on an IA64 architecture is the pointer going to be 8 bytes aligned ? Yes, kmalloc results are always 'sizeof(void*)' aligned. Do you have stricter alignment requirements? -- Manfred - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: Q: How do I get from the latest stable kernel version to the latest prepatch version ?
Thanks. It just struck me odd that the latest is 2.4.2 while the prepatches were 2.4.3 so I figured there must be something I missed in between (my logic told me that a 2.4.3 patch would be against a 2.4.3 something ;-). BTW, I haven't seen any announcements from Linus in this mailing list regarding new versions, just the updates on the web site and Alan's release notes saying he's merging with 2.4.3xx. Are those announcements being posted somewhere else now ? -Original Message- From: Leonid Mamtchenkov [mailto:[EMAIL PROTECTED]] Sent: Monday, March 26, 2001 2:33 PM To: Hen, Shmulik Cc: 'LKML' Subject: Re: Q: How do I get from the latest stable kernel version to the late st prepatch version ? Hello Hen, Shmulik, Once you wrote about "Q: How do I get from the latest stable kernel version to the late st prepatch version ?": HS According to http://www.kernel.org, the latest stable kernel version is HS 2.4.2. The latest prepatch version is 2.4.3-pre3. HS HS In order to get a full 2.4.3-pre8 kernel do I have to: HS HS A. download linux-2.4.2.tar.gz and all the patch-2.4.3-preX.gz and apply HS them in succession or, HS B. download linux-2.4.3.tar.gz (exists ?) and then apply the all patches or, HS C. download linux-2.4.3-pre7.tar.gz (exists ?) and apply only HS patch-2.4.3-pre8.gz ? Download 2.4.2 and then apply 2.4.3-preX (latest) on it... that's it. You might want to visit http://kernelnewbies.org . They have some good docs there. -- Best regards, Leonid Mamtchenkov System Administrator - 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/
RE: URGENT : System hands on Freeing unused kernel memory:
Does it hang forever ? I've noticed that my kernel (2.4.2) stalls for several minutes with the same message but suddenly after that the login prompt appears (anything between, like configurations and services starting messages, are gone). We've been able to track it down to a change we did to /etc/lilo.conf to add support for kernel prints to go out to a serial debugger. Before that everything was OK, but after we added append="console=tty0 console=ttyS1,38400", this problem started. We did notice however that everything that doesn't appear on the console does appear on the serial debugger. Shmulik. -Original Message- From: Thomas Foerster [mailto:[EMAIL PROTECTED]] Sent: Tuesday, March 27, 2001 10:40 AM To: [EMAIL PROTECTED] Subject: Re: URGENT : System hands on "Freeing unused kernel memory: " On 03.27 Thomas Foerster wrote: But suddenly the box was offline. One technical assistant from our ISP tried to reboot our server (he couldn't tell me if there had been any messages on the screen), but the system always hangs on Freeing unused kernel memory: xxk freed Try booting with init=/bin/bash, it looks like kernel gets a bad /sbin/init, and gets stuck. Perhaps the shutdown damaged init, it starts to run and get hung. That didn't fix the problem :( When i run "diff" on a new and the "old" init, i get no diffs ... Must be something other :( Thomas - 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/ - 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/
RE: Plans for 2.5
Just some general questions: 1) Is there anywhere a list that describes what is intended to be in 2.5.x ? 2) Are there any early releases of 2.5.x ? 3) Are the things for 2.5.x being discussed on another mailing list ? 4) What is the time frame of releasing 2.5.x-final (or 2.6.x) ? Specifically, I'm more interested in the network driver aspect. 1) Are there any intended changes to the networking layer ? 2) I over heard something about making the driver reentrant - any news ? 3) What about support for IPv6 ? (I noticed it was marked as experimental until now) Thanks in advance, Shmulik Hen Software Engineer Linux Advanced Networking Services Intel Network Communications Group Jerusalem, Israel -Original Message- From: Bruno Avila [mailto:[EMAIL PROTECTED]] Sent: Thursday, March 29, 2001 12:45 AM To: [EMAIL PROTECTED] Subject: Plans for 2.5 Hello people, I got some questions. When are we going to develop stuff for 2.5? What is planed? My opinion for linux 2.5 should be performance. Since linux already is stable or well done for nature, we could thing more on performance to be a diferencial over others. What do you people thing? Bruno Avila PS: Not a good english. I know! :) - 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/ - 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/
change_mtu boundary checking error
Hello, Going through the change_mtu() code in the kernel, I came across the default function supplied when calling ether_setup(). I could see that eth_change_mtu() (drivers/net/net_init.c) does the following: if( (new_mtu 68) || (new_mtu 1500) ) return -EINVAL; Looking in include/linux/if_ether.h I found the following constants: #define ETH_ALEN6 /* Octets in one ethernet addr */ #define ETH_HLAN14 /* Total octets in header. */ #define ETH_ZLEN60 /* Min. octets in frame sans FCS */ #define ETH_DATA_LEN1500/* Max. octets in payload */ #define ETH_FRAME_LEN 1514/* Max. octets in frame sans FCS */ Now, the high boundary seemed reasonable (ETH_FRAME_LEN - ETH_HLEN = ETH_DATA_LEN) which gives 1500, but why is the low boundary set to 68 ? According to my calculations, it should have been ETH_ZLEN - ETH_HLEN which gives 46. Doesn't mtu means only the payload size ? Where did the 68 come from ? Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) - 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/
RE: ioctl call for network device
> struct ifreq has a member called ifr_data. It is a pointer. You can > put a pointer to any of your data, including the most complex structure > you might envision, in that area. This allows you to pass anything > to and from your module. This pointer can be properly dereferenced > in kernel space but you should use copy_to/from_user and friends so a > user-space coding bug won't panic the kernel. How about a linked list ? Will the driver be able to follow the list where each node was dynamically allocated by the application ? Is there a size limit on the buffer ifr_data points to ? (AFAIK, Windows NDIS drivers limit to 1 page buffer =4096 bytes). Thanks, Shmulik Hen Linux Advanced Networking Services Intel Network Communications Group - 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/
RE: spinlock help
How about if the same sequence occurred, but from two different drivers ?
We've had some bad experience with this stuff. Our driver, which acts as an
intermediate net driver, would call the hard_start_xmit in the base driver.
The base driver, wanting to block receive interrupts would issue a
'spin_lock_irqsave(a,b)' and process the packet. If the TX queue is full, it
could call an indication entry point in our intermediate driver to signal it
to stop sending more packets. Since our indication function handles many
types of indications but can process them only one at a time, we wanted to
block other indications while queuing the request.
The whole sequence would look like that:
[our driver]
ans_send() {
.
.
e100_hard_start_xmit(dev, skb);
.
.
}
[e100.o]
e100_hard_start_xmit() {
.
.
spin_lock_irqsave(a,b);
.
.
if(tx_queue_full)
ans_notify(TX_QUEUE_FULL); <--
.
.
spin_unlock_irqrestore(a,b);
}
[our driver]
ans_notify() {
.
.
spin_lock_irqsave(c,d);
queue_request(req_type);
spin_unlock_irqrestore(c,d);<--
.
.
}
At that point, for some reason, interrupts were back and the e100.o would
hang in an infinite loop (we verified it on kernel 2.4.0-test10 +kdb that
the processor was enabling interrupts and that the e100_isr was called for
processing an Rx int.).
How is that possible that a 'spin_unlock_irqrestore(c,d)' would also restore
what should have been restored only with a 'spin_unlock_irqrestore(a,b)' ?
Thanks in advance,
Shmulik Hen
Software Engineer
Linux Advanced Networking Services
Intel Network Communications Group
Jerusalem, Israel.
-Original Message-
From: Nigel Gamble [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 1:54 AM
To: Manoj Sontakke
Cc: [EMAIL PROTECTED]
Subject: Re: spinlock help
On Tue, 6 Mar 2001, Manoj Sontakke wrote:
> 1. when spin_lock_irqsave() function is called the subsequent code is
> executed untill spin_unloc_irqrestore()is called. is this right?
Yes. The protected code will not be interrupted, or simultaneously
executed by another CPU.
> 2. is this sequence valid?
> spin_lock_irqsave(a,b);
> spin_lock_irqsave(c,d);
Yes, as long as it is followed by:
spin_unlock_irqrestore(c, d);
spin_unlock_irqrestore(a, b);
Nigel Gamble[EMAIL PROTECTED]
Mountain View, CA, USA. http://www.nrg.org/
-
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/
-
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/
RE: spinlock help
spin_lock_bh() won't block interrupts and we need them blocked to prevent
more indications.
spin_lock_irq() could do the trick but it's counterpart spin_unlock_irq()
enables all interrupts by calling sti(), and this is even worse for us.
Shmulik.
-Original Message-
From: Manoj Sontakke [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 12:27 PM
To: Hen, Shmulik
Cc: '[EMAIL PROTECTED]'; Manoj Sontakke; [EMAIL PROTECTED]
Subject: Re: spinlock help
hi
spin_lock_irq() andspin_lock_bh()
can they be of any use to u?
"Hen, Shmulik" wrote:
>
> How about if the same sequence occurred, but from two different drivers ?
>
> We've had some bad experience with this stuff. Our driver, which acts as
an
> intermediate net driver, would call the hard_start_xmit in the base
driver.
> The base driver, wanting to block receive interrupts would issue a
> 'spin_lock_irqsave(a,b)' and process the packet. If the TX queue is full,
it
> could call an indication entry point in our intermediate driver to signal
it
> to stop sending more packets. Since our indication function handles many
> types of indications but can process them only one at a time, we wanted to
> block other indications while queuing the request.
>
> The whole sequence would look like that:
>
> [our driver]
> ans_send() {
> .
> .
> e100_hard_start_xmit(dev, skb);
> .
> .
> }
>
> [e100.o]
> e100_hard_start_xmit() {
> .
> .
> spin_lock_irqsave(a,b);
> .
> .
> if(tx_queue_full)
> ans_notify(TX_QUEUE_FULL); <--
> .
> .
> spin_unlock_irqrestore(a,b);
> }
>
> [our driver]
> ans_notify() {
> .
> .
> spin_lock_irqsave(c,d);
> queue_request(req_type);
> spin_unlock_irqrestore(c,d);<--
> .
> .
> }
>
> At that point, for some reason, interrupts were back and the e100.o would
> hang in an infinite loop (we verified it on kernel 2.4.0-test10 +kdb that
> the processor was enabling interrupts and that the e100_isr was called for
> processing an Rx int.).
>
> How is that possible that a 'spin_unlock_irqrestore(c,d)' would also
restore
> what should have been restored only with a 'spin_unlock_irqrestore(a,b)' ?
>
> Thanks in advance,
> Shmulik Hen
> Software Engineer
> Linux Advanced Networking Services
> Intel Network Communications Group
> Jerusalem, Israel.
>
> -Original Message-
> From: Nigel Gamble [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, March 07, 2001 1:54 AM
> To: Manoj Sontakke
> Cc: [EMAIL PROTECTED]
> Subject: Re: spinlock help
>
> On Tue, 6 Mar 2001, Manoj Sontakke wrote:
> > 1. when spin_lock_irqsave() function is called the subsequent code is
> > executed untill spin_unloc_irqrestore()is called. is this right?
>
> Yes. The protected code will not be interrupted, or simultaneously
> executed by another CPU.
>
> > 2. is this sequence valid?
> > spin_lock_irqsave(a,b);
> > spin_lock_irqsave(c,d);
>
> Yes, as long as it is followed by:
>
> spin_unlock_irqrestore(c, d);
> spin_unlock_irqrestore(a, b);
>
> Nigel Gamble[EMAIL PROTECTED]
> Mountain View, CA, USA. http://www.nrg.org/
>
> -
> 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/
>
> -
> 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/
--
Regards,
Manoj Sontakke
-
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/
RE: spinlock help
e100 implements all sorts of hooks for our intermediate driver (kind of a
co-development effort), so eepro100 is out of the question for us.
Shmulik.
-Original Message-
From: Ofer Fryman [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 12:31 PM
To: 'Hen, Shmulik'
Cc: [EMAIL PROTECTED]
Subject: RE: spinlock help
Did you try looking at Becker eepro100 driver it seems to be simple, no
unnecessary spin_lock_irqsave?.
-Original Message-
From: Hen, Shmulik [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 11:21 AM
To: '[EMAIL PROTECTED]'; Manoj Sontakke
Cc: [EMAIL PROTECTED]
Subject: RE: spinlock help
How about if the same sequence occurred, but from two different drivers ?
We've had some bad experience with this stuff. Our driver, which acts as an
intermediate net driver, would call the hard_start_xmit in the base driver.
The base driver, wanting to block receive interrupts would issue a
'spin_lock_irqsave(a,b)' and process the packet. If the TX queue is full, it
could call an indication entry point in our intermediate driver to signal it
to stop sending more packets. Since our indication function handles many
types of indications but can process them only one at a time, we wanted to
block other indications while queuing the request.
The whole sequence would look like that:
[our driver]
ans_send() {
.
.
e100_hard_start_xmit(dev, skb);
.
.
}
[e100.o]
e100_hard_start_xmit() {
.
.
spin_lock_irqsave(a,b);
.
.
if(tx_queue_full)
ans_notify(TX_QUEUE_FULL); <--
.
.
spin_unlock_irqrestore(a,b);
}
[our driver]
ans_notify() {
.
.
spin_lock_irqsave(c,d);
queue_request(req_type);
spin_unlock_irqrestore(c,d);<--
.
.
}
At that point, for some reason, interrupts were back and the e100.o would
hang in an infinite loop (we verified it on kernel 2.4.0-test10 +kdb that
the processor was enabling interrupts and that the e100_isr was called for
processing an Rx int.).
How is that possible that a 'spin_unlock_irqrestore(c,d)' would also restore
what should have been restored only with a 'spin_unlock_irqrestore(a,b)' ?
Thanks in advance,
Shmulik Hen
Software Engineer
Linux Advanced Networking Services
Intel Network Communications Group
Jerusalem, Israel.
-Original Message-
From: Nigel Gamble [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 1:54 AM
To: Manoj Sontakke
Cc: [EMAIL PROTECTED]
Subject: Re: spinlock help
On Tue, 6 Mar 2001, Manoj Sontakke wrote:
> 1. when spin_lock_irqsave() function is called the subsequent code is
> executed untill spin_unloc_irqrestore()is called. is this right?
Yes. The protected code will not be interrupted, or simultaneously
executed by another CPU.
> 2. is this sequence valid?
> spin_lock_irqsave(a,b);
> spin_lock_irqsave(c,d);
Yes, as long as it is followed by:
spin_unlock_irqrestore(c, d);
spin_unlock_irqrestore(a, b);
Nigel Gamble[EMAIL PROTECTED]
Mountain View, CA, USA. http://www.nrg.org/
-
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/
-
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/
-
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/
RE: spinlock help
The kdb trace was accurate, we could actually see the e100 ISR pop from no
where right in the middle of our ans_notify every time the TX queue would
fill up. When we commented out the call to spin_*_irqsave(), it worked fine
under heavy stress for days.
Is it possible it was something wrong with 2.4.0-test10 specifically ?
We had to drop the locks in the final release and never got around to
checking it on other kernel releases (it went on the TO_DO list ;-).
Shmulik.
-Original Message-
From: Andrew Morton [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 07, 2001 1:43 PM
To: Hen, Shmulik
Subject: Re: spinlock help
"Hen, Shmulik" wrote:
>
> How about if the same sequence occurred, but from two different drivers ?
>
> We've had some bad experience with this stuff. Our driver, which acts as
an
> intermediate net driver, would call the hard_start_xmit in the base
driver.
> The base driver, wanting to block receive interrupts would issue a
> 'spin_lock_irqsave(a,b)' and process the packet. If the TX queue is full,
it
> could call an indication entry point in our intermediate driver to signal
it
> to stop sending more packets. Since our indication function handles many
> types of indications but can process them only one at a time, we wanted to
> block other indications while queuing the request.
>
> The whole sequence would look like that:
>
> [our driver]
> ans_send() {
> .
> .
> e100_hard_start_xmit(dev, skb);
> .
> .
> }
>
> [e100.o]
> e100_hard_start_xmit() {
> .
> .
> spin_lock_irqsave(a,b);
> .
> .
> if(tx_queue_full)
> ans_notify(TX_QUEUE_FULL); <--
> .
> .
> spin_unlock_irqrestore(a,b);
> }
>
> [our driver]
> ans_notify() {
> .
> .
> spin_lock_irqsave(c,d);
> queue_request(req_type);
> spin_unlock_irqrestore(c,d);<--
> .
> .
> }
>
> At that point, for some reason, interrupts were back
Sorry, that can't happen.
Really, you must have made a mistake somewhere.
-
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/
RE: spinlock help
OK guys, you were right. The bug was in our code - sorry for trouble. Turns out that while I was away, the problem was solved by someone else. The problem is probably related to the fact that when we did 'spin_lock_irqsave(c,d)', 'd' was a global variable. The fix was to wrap the call with another function and declare 'd' as local. I can't quite explain, but I think that changing from a static to automatic variable made the difference. My best guess is that since 'd' is passed by value and not by reference, the macro expansion of spin_lock_irqsave() relies on the location of 'd' in the stack and if 'd' was on the heap instead, it might get trashed. I would really like to hear your expert opinion on my assumption. Thanks, Shmulik. -Original Message- From: Andrew Morton [mailto:[EMAIL PROTECTED]] Sent: Wednesday, March 07, 2001 2:54 PM To: Hen, Shmulik Cc: 'LKML' Subject: Re: spinlock help "Hen, Shmulik" wrote: > > The kdb trace was accurate, we could actually see the e100 ISR pop from no > where right in the middle of our ans_notify every time the TX queue would > fill up. When we commented out the call to spin_*_irqsave(), it worked fine > under heavy stress for days. > > Is it possible it was something wrong with 2.4.0-test10 specifically ? > Sorry, no. If spin_lock_irqsave()/spin_unlock_irqrestore() were accidentally reenabling interrupts then it would be the biggest, ugliest catastrophe since someone put out a kernel which forgot to flush dirty inodes to disk :) Conceivably it was a compiler bug. Were you using egcs-1.1.2/x86? - - 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/
Q: network drivers interface changes
Hello, Is there a good source of information that describes the changes in network driver interface between 2.2.x and 2.4.x kernels ? Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Intel Network Communications Group Jerusalem - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
[OT] ethtool MII helpers (actually two OT's)
MII --- Is there any support in the MII standard for 1000Mbps (GbE Fiber/Copper) ? Perhaps an extension to the standard ? I could see that some of the Gigabit adapters supported by the kernel provide the MII IOCTLs interface, but couldn't figure out how to extract the correct speed information from the registers I can read. I know it's a bit of a hassle and I have to get the local capabilities and match them against the partner's capabilities and find the highest common speed etc. etc. but I'm sure that if the driver can do it I can reproduce it in userland too. EthTool --- Is there a way that I can extract the link status information out of the ethtool struct ? I could see that at least one Gigabit adapter driver (bcm5700.c), provides the EthTool interface and reports the correct speed and duplex mode but not the link status. Is there a place that defines how a driver is supposed to implement the support for EthTool ? I figured that since there is no separate field for link status (at least in version 1.2), a driver is supposed to report speed=0 or something like that when the link is down. I know this driver detects link status changes for sure because it prints messages every time, but the speed and duplex are always reported the same. Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Intel Network Communications Group Jerusalem, Israel -Original Message- From: Jeff Garzik [mailto:[EMAIL PROTECTED]] Sent: Friday, June 22, 2001 8:59 AM To: Chris Wedgwood Cc: Linux Kernel Mailing List; [EMAIL PROTECTED]; David S. Miller Subject: Re: PATCH: ethtool MII helpers Chris Wedgwood wrote: > > On Fri, Jun 22, 2001 at 01:24:36AM -0400, Jeff Garzik wrote: > > Sure, and that's planned. Wanna send me a patch for it? :) > > Possibly, but I wonder if this is a kernel-space problem or not. Why > not put all the smarts into userland for it? I meant, send me a patch for userland ethtool, to do exactly what you described. > It will definitely fall back on the MII ioctls if ethtool media > support for the desired command doesn't exist. > > Well, that is more or less as much as needs to be done. That, and > some kind of super-set API to be defined for all new stuff, having > two slightly different APIs for the same things sucks. Both APIs do different things but have a common subset, yes. The MII ioctls only do their thing for MII-like hardware. ethtool can be applied to any hardware. Old ISA drivers that don't do MII, or do it in a really nonstandard way. For example I have ethtool code locally which allows ne2k-pci to do media selection via ioctl, for two popular ne2k cards, something its never been able to do before. Emulating media selection support for things like 10base2<->10baseT<->AUI just isn't possible with the MII ioctls. MII is a standard and incredibly popular, thus mii-tool works most popular PCI NICs, for the most popular media types. But it's still basically a hardware interface. I am not convinced its a good idea for make the [G]MII ioctls the Linux software media interface for all network hardware. I see ethtool as the interface for tuning your NIC, that works across all hardware. I see mii-diag as the way to do advance MII-specific hardware stuff, like next page or HA monitoring or whatever. Jeff -- Jeff Garzik | Andre the Giant has a posse. Building 1024| MandrakeSoft | - 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/ - 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/
catch 22 - porting net driver from 2.2 to 2.4
Hello, This is a bit long and I apologize (since there are kdb captures in it). We are developing an advanced networking services driver (loadable module) and are having problems porting it to work on 2.4.x kernel. The driver is supposed to provide services such as fault tolerance, load balancing, link aggregation and VLAN. It does that by creating a group of "virtual" adapters that are bound on top of a team of network adapters. This works great on 2.2 kernels but demonstrated a few problems on 2.4.0-test9 The only problem we have left now has to do with insmod/rmmod. for good reasons, we cant just call init_etherdev() like base drivers do, so we created our own version that handles memory and name allocations and calls register_netdevice() on it's own the same as init_etherdev(). since we've got several "virtual" adapters that are part of a topology being built progressively, we can't perform their registrations during module_init() but rather through an IOCTL and here is our problem: if we call register_netdevice(), we get the following message: RTNL: assertion failed at devinet.c(775):inetdev_event we figured this is because we neglected rtnl_lock() so instead we try using register_netdev() to handle this for us but then we get: Scheduling in interrupt kernel BUG at sched.c:696! Entering kdb (current=0xc51a8000, pid 1075) on processor 1 Panic: invalid operand due to panic @ 0xc011aa71 eax = 0x001b ebx = 0x0020 ecx = 0xc030d80c edx = 0x esi = 0xc030b5b4 edi = 0x esp = 0xc51a9d34 eip = 0xc011aa71 ebp = 0xc51a9d8c ss = 0x0018 cs = 0x0010 eflags = 0x00010246 ds = 0x0018 es = 0x0018 origeax = 0x = 0xc51a9d00 [1]kdb> bt EBP EIP Function(args) 0xc51a9d8c 0xc011aa71 schedule+0x935 (0xc2c9e000, 0xc4482520, 0xc51a9e10) kernel .text 0xc010 0xc011a13c 0xc011aa80 0xc51a9db8 0xc0107b8d __down+0xf5 kernel .text 0xc010 0xc0107a98 0xc0107c68 0xc51a9dcc 0xc0107f43 __down_failed+0xb (0xc51a9de4, 0xd082de5c, 0xc2c9e000, 0xc60a8320, 0xc51a9dfc) kernel .text 0xc010 0xc0107f38 0xc0107f4c 0xc023a7a9 stext_lock+0x4919 kernel .text.lock 0xc0235e90 0xc0235e90 0xc023bd80 0xc51a9dd4 0xc01f2e81 rtnl_lock+0x11 (0xc2c9e000) kernel .text 0xc010 0xc01f2e70 0xc01f2e88 0xc51a9de4 0xd082de5c [ians]iansInitEtherdev+0x20 (0xc4482520) ians .text 0xd082d060 0xd082de3c 0xd082de78 . . (boring chain of calls) . 0xc51a9ec4 0xd082dbcd [ians]doControlIoctl+0x15d (0xc2c9e200, 0xc51a9f20, 0x89f0) ians .text 0xd082d060 0xd082da70 0xd082dc40 0xc51a9ee4 0xc01ef09f dev_ifsioc+0x33f (0xc51a9f20, 0x89f0, 0xc51a9f20) kernel .text 0xc010 0xc01eed60 0xc01ef0b0 0xc51a9f40 0xc01ef29d dev_ioctl+0x1ed (0x89f0, 0xba58) kernel .text 0xc010 0xc01ef0b0 0xc01ef300 0xc51a9f64 0xc021a70c inet_ioctl+0x18c (0xc339d13c, 0x89f0, 0xba58) kernel .text 0xc010 0xc021a580 0xc021a720 0xc51a9f84 0xc01e8f06 sock_ioctl+0x9e (0xc339d040, 0xc38e0900, 0x89f0, 0xba58) kernel .text 0xc010 0xc01e8e68 0xc01e8f6c 0xc51a9fbc 0xc014f5fd sys_ioctl+0x26d (0x3, 0x89f0, 0xba58, 0x4000ae60, 0xbba4) kernel .text 0xc010 0xc014f390 0xc014f6a0 0xc010965f system_call+0x33 kernel .text 0xc010 0xc010962c 0xc0109664 We figured that since we are in user context (do_ioctl) and use spin_lock_bh() to protect us from other concurrent threads, it might interfere with rtnl_lock() so we remove our lock just before calling register_netdev() and lock again upon return but then the whole process just stopped and didn't return to the prompt. from within kdb, we could see that all CPU's are running in idle but if we try to return to the prompt the whole system hangs. sometimes it hangs if we try to run ifconfig -a to see if the virtual adapters appear. I can't use it without locks, I can't use it with locks and I can't complete the operation if I remove my own locks - catch 22. The other problem has to do with rmmod - here we get called in our cleanup_module function and from it we try to call unregister_netdev() for each registered virtual adapter. in this case, we get: Scheduling in interrupt kernel BUG at sched.c:696! Entering kdb (current=0xc38c8000, pid 1602) on processor 0 Panic:
Q: kallsyms - where can I find it and what does it do ?
Hello, I'm trying to build a new kernel with kdb support and I keep getting an error from the make file: kallsyms pass 1 [make] /bin/sh: /sbin/kallsyms: No such file or directory error What is kallsyms and where can I get it from ? Here is what I have and what I did: The machine is a Compaq Ap500 dual P-III Xeon. I'm using RedHat 6.2 (clean disk - custom install - install everything). The new kernel is linux-2.4.0-test9.tar.gz (the latest according to www.kernel.org). The kdb patch is kdb-v1.5-2.4.0-test9-pre9.gz (the latest according to oss.sgi.com). There doesn't seem to be a problem running the patch (no error messages, at least). I made sure kdb support is checked under 'make menuconfig'. I ran 'make dep; make clean; make bzImage' and it keeps failing. While on the subject, how is it possible that I can build an SMP kernel on that machine (before the patch), but can't build a UP kernel ?. I use 'make menuconfig' and uncheck "SMP support", then run 'make dep; make clean; make bzImage' and I get all kinds of warnings and errors until the make file simply stops running (doesn't return to the prompt with an error message - just stops and I have to hit to get back to the prompt). Thanks in advance, Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Please ignore - RE: kallsyms - where can I find it and what does it do ?
-Original Message- From: Hen, Shmulik [mailto:[EMAIL PROTECTED]] Sent: Wednesday, October 18, 2000 6:17 PM To: [EMAIL PROTECTED] Cc: 'Keith Owens' Subject: Q: kallsyms - where can I find it and what does it do ? Hello, I'm trying to build a new kernel with kdb support and I keep getting an error from the make file: kallsyms pass 1 [make] /bin/sh: /sbin/kallsyms: No such file or directory error What is kallsyms and where can I get it from ? Here is what I have and what I did: The machine is a Compaq Ap500 dual P-III Xeon. I'm using RedHat 6.2 (clean disk - custom install - install everything). The new kernel is linux-2.4.0-test9.tar.gz (the latest according to www.kernel.org). The kdb patch is kdb-v1.5-2.4.0-test9-pre9.gz (the latest according to oss.sgi.com). There doesn't seem to be a problem running the patch (no error messages, at least). I made sure kdb support is checked under 'make menuconfig'. I ran 'make dep; make clean; make bzImage' and it keeps failing. While on the subject, how is it possible that I can build an SMP kernel on that machine (before the patch), but can't build a UP kernel ?. I use 'make menuconfig' and uncheck "SMP support", then run 'make dep; make clean; make bzImage' and I get all kinds of warnings and errors until the make file simply stops running (doesn't return to the prompt with an error message - just stops and I have to hit to get back to the prompt). Thanks in advance, Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/ - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
page fault problems porting a network driver to 2.4.x
Hello,
We are developing an advanced networking services loadable module and are
having problems porting it to work on 2.4.x kernels. The driver is supposed
to provide services such as fault tolerance, load balancing and link
aggregation over a team of network adapters. It works OK on 2.2.x kernels
but hangs on 2.4.x kernels.
In order to debug it, we stripped it down to become a mere "intermediate" or
"filter" driver that binds to a base driver and passes everything through in
both directions (Rx, Tx, IOCTL, stats, etc.). After going through the basics
of modifying the driver to compile on 2.4.x kernels and fighting some nasty
dead locks due to the new nature of the networking layer, we managed to get
it to run. The driver will receive and transmit a few hundreds of thousands
of packets (while having a periodic timer expire 10 times a second and
running continuous IOCTLs), and then it causes an oops about not being able
to handle a page fault.
The function looks something like:
int iansHardStartXmit(struct sk_buff *skb, struct net_device *dev) {
int res;
struct net_device *base;
spin_lock();
base = get_base_driver_by_name(name);
if(base != NULL) {
res = base->hard_start_xmit(skb, base);
}
spin_unlock();
return res;
}
We used kdb in order to track down the problem and found out the following
stack trace:
EBPEIP function(args)
0xc4cd1c54 0xd081e3e7 [e100]__kallsyms+0xb (0xc4b595a0,
0xc840f200)
e100 __kallsyms 0xd081e3dc
0xd081e3dc 0xd0820dsc
0xd08244ba [ians]iansHardStartXmit+0xa6 (0xc4b595a0,
0xc4d9bc00)
ians .text 0xd0824060 0xd0824414
0xd082452c
0xc01f9d1f qdisc_restart+0xcf (0xc4d9bc00)
kernel .text 0xc010 0xc01f9c50
0xc01f9f14
*
*
*
This goes on and shows that this is an ICMP echo reply packet going down
through the IP stack to the filter driver (apparently 0xc4b595a0 is the skb,
0xc4d9bc00 is the *dev of the filter driver and 0xc840f200 is the *dev of
the base driver). The filter driver is supposed to call the
dev->hard_start_xmit of the base driver, but strangely it lands somewhere in
the data segment of the base driver (__kallsyms is a part of the symbol
table of the module according to insmod -m).
Figuring the dev->hard_start_xmit pointer got trashed somehow, we added a
check to make sure the same pointer is always called, and indeed this was
the case. Looking at the assembly code with kdb, we could see that the call
to the base driver is done by a 'call *%eax' command. kdb reports that
eax=0x after the page fault (origeax).
How is it possible that the pointer to the function keeps it's value, but
the jump to that function falls somewhere else ?
The entire function is protected by a spinlock, so there is no worry about
the other threads messing my data.
We are using:
RedHat 6.2
gcc v2.91.66
modutils v2.3.11-1
kernel linux-2.4.0-test9
kdb v1.5-2.4.0-test9-pre9
Compaq ap500 dual p-III Xeon
Thanks,
Shmulik Hen
Software Engineer
Linux Advanced Networking Services
Network Communications Group, Israel (NCGj)
Intel Corporation Ltd.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
Please read the FAQ at http://www.tux.org/lkml/
pointer to dev->hard_start_xmit() gets trashed in 2.4.0-test9
Hello,
We are developing an advanced networking services loadable module and are
having problems porting it to work on 2.4.x kernels. The driver is supposed
to provide services such as fault tolerance, load balancing and link
aggregation over a team of network adapters. It works OK on 2.2.x kernels
but hangs on 2.4.x kernels.
In order to debug it, we stripped it down to become a mere "intermediate" or
"filter" driver that binds to a base driver and passes everything through in
both directions (Rx, Tx, IOCTL, stats, etc.). After going through the basics
of modifying the driver to compile on 2.4.x kernels and fighting some nasty
dead locks due to the new nature of the networking layer, we managed to get
it to run. The driver will receive and transmit a few hundreds of thousands
of packets (while having a periodic timer expire 10 times a second and
running continuous IOCTLs), and then it causes an oops about not being able
to handle a page fault.
The function looks something like:
int iansHardStartXmit(struct sk_buff *skb, struct net_device *dev) {
int res;
struct net_device *base;
spin_lock(); //no interrupts involved, so spin_lock
should do
base = get_base_driver_by_name(name);
if(base != NULL) {
BUG_TRAP(ptr_g == base->hard_start_xmit); //make sure it's
always the same addr
res = base->hard_start_xmit(skb, base);
}
spin_unlock();
return res;
}
We used kdb in order to track down the problem and found out the following
stack trace:
EBPEIP function(args)
0xc4cd1c54 0xd081e3e7 [e100]__kallsyms+0xb (0xc4b595a0,
0xc840f200)
e100 __kallsyms 0xd081e3dc
0xd081e3dc 0xd0820dsc
0xd08244ba [ians]iansHardStartXmit+0xa6 (0xc4b595a0,
0xc4d9bc00)
ians .text 0xd0824060 0xd0824414
0xd082452c
0xc01f9d1f qdisc_restart+0xcf (0xc4d9bc00)
kernel .text 0xc010 0xc01f9c50
0xc01f9f14
*
*
*
This goes on and shows that this is an ICMP echo reply packet going down
through the IP stack to the filter driver (apparently 0xc4b595a0 is the skb,
0xc4d9bc00 is the *dev of the filter driver and 0xc840f200 is the *dev of
the base driver). The filter driver is supposed to call the
dev->hard_start_xmit of the base driver, but strangely it lands somewhere in
the data segment of the base driver (__kallsyms is a part of the symbol
table of the module according to insmod -m).
Figuring the dev->hard_start_xmit pointer got trashed somehow, we added a
check to make sure the same pointer is always called, and indeed this is the
case. Looking at the assembly code with kdb, we could see that the call to
the base driver is done by a 'call *%eax' instruction.
How is it possible that the pointer to the function keeps it's value, but
the jump to that function falls somewhere else ?
We are using:
RedHat 6.2
gcc v2.91.66
modutils v2.3.11-1 (was upgraded because of kdb)
kernel linux-2.4.0-test8 (SMP, +kdb, compiled with frame pointers,
SPINLOCK_DEBUG=2)
kdb v1.4-2.4.0-test9-pre9
Compaq ap500 dual p-III Xeon
Could this be a version mismatch between the components above ?
Thanks,
Shmulik Hen
Software Engineer
Linux Advanced Networking Services
Network Communications Group, Israel (NCGj)
Intel Corporation Ltd.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
Please read the FAQ at http://www.tux.org/lkml/
Multiple warnings when compiling network driver in 2.4.0-test9
Hello,
While trying to compile a network driver for 2.4.0-test9 (+kdb-v1.5,
configured for UP) I'm getting multiple warnings:
/usr/src/linux/include/linux/sched.h:700: warning: can't inline call
to `__mmdrop'
/usr/src/linux/include/linux/sched.h:704: warning: called from here
This happens for every kernel header I try to use such as netdevice.h,
skbuff.h, malloc.h and pci.h
Each of those header files includes slab.h (line 14) that includes mm.h
(line 4) that includes sched.h which contains the following on line 700:
700:extern inline void FASTCALL(__mmdrop(struct mm_struct *));
701:static inline void mmdrop(struct mm_struct * mm)
702:{
703:if (atomic_dec_and_test(>mm_count))
704:__mmdrop(mm);
705:}
My make file uses the following flags:
gcc -fomit-frame-pointer -Wall -Wstrict-prototypes -Winline -O3
-D__KERNEL__ -DMODULE -DDEBUG -DMODVERSIONS -I/usr/src/linux/include
Can anyone tell me what this warning means and if I can safely ignore it (or
expect disaster) ?
Thanks,
Shmulik Hen,
Software Engineer
Linux Advanced Networking Services
Network Communications Group, Israel (NCGj)
Intel Corporation Ltd.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
Please read the FAQ at http://www.tux.org/lkml/
Locking Between User Context and Soft IRQs in 2.4.0
Hello, We are trying to port a network driver from 2.2.x to 2.4.x and have some question regarding locks. According to the kernel locking HOWTO, we have to take extra care when locking between user context threads and BH/tasklet/softIRQ, so we learned (the hard way ;-) that when running the ioctl system call from an application we should use spin_lock/unlock_bh() and not spin_lock/unlock() inside dev->do_ioctl(). * What about the other entry points implemented in net_device ? * We've got dev->get_stats, dev->set_mac_address, dev->set_mutlicast_list and others that are all called from running 'ifconfig' which is an application. Are they considered user context too ? * What about dev->open and dev->stop ? * We figured that dev->hard_start_xmit() and timer callbacks are not considered user context, but how can I find out if they are being run as SoftIRQ or as tasklets or as Bottom Halves ? (their different definitions require different types of protections) Our driver is actually an intermediate driver bound on top of a regular net driver. It behaves both as a network adapter driver and a protocol at the same time. I can safely assume that it will have to handle both transmits and receives simultaneously (no hardware interrupts are involved). We've decided that for the first stage we are going to implement "wide" locks that wrap entire operations from top to bottom. For example, our dev->hard_start_xmit() will have a spin_lock() at the beginning and a spin_unlock() at the end of the function. * Will it be safe to keep the lock until after the call to the base driver's hard_start_xmit, or do I have to release the lock just before that ? * Or, in our receive function, will I have to release the lock before or after the call to netif_rx() ? * What about other calls to the kernel ? can the running thread be switched out of context when calling kernel entries and not be switched back in when they finish ? should I beware of deadlocks in such case ? Thanks in advance, Shmulik Hen, Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: catch 22 - porting net driver from 2.2 to 2.4
and you don't get the "RTNL: assertion failed at devinet.c(775):inetdev_event" in 2.4.x ? the thing is I need to prevent Tx/Rx when a topology change is initiated from the ioctl (registering a virtual adapter is just one example), so they all share a single lock and I must use spin_lock_bh from the ioctl. Shmulik. -Original Message- From: Olaf Titz [mailto:[EMAIL PROTECTED]] Sent: Friday, November 10, 2000 2:09 AM To: [EMAIL PROTECTED] Subject: Re: catch 22 - porting net driver from 2.2 to 2.4 > We figured that since we are in user context (do_ioctl) and use > spin_lock_bh() to protect us from other concurrent threads, it might > interfere with rtnl_lock() so we remove our lock just before calling > register_netdev() and lock again upon return but then the whole process just > stopped and didn't return to the prompt. from within kdb, we could see that Can't you just do this: #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,3,0) /* not sure about the 0 */ #define rtnl_LOCK() rtnl_lock() #define rtnl_UNLOCK() rtnl_unlock() #else #define rtnl_LOCK() /* nop */ #define rtnl_UNLOCK() /* nop */ #endif rtnl_LOCK(); register_netdevice(...); rtnl_UNLOCK(); that works for me (yes, from do_ioctl, but without the bh lock - I don't know if that's absolutely needed in your case). Olaf - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/ - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: catch 22 - porting net driver from 2.2 to 2.4
So how come I get the "RTNL: assertion failed at devinet.c(775):inetdev_event" when I call register_netdevice without rtnl_lock/unlock ? could it be a 2.4.0-test9 thing ? (haven't used test10 or 11 yet). and what about rmmod causing the panic when I use unregister_netdev or never completing the operation when I use unregister_netdevice ? does module_exit run inside rtnl_lock too ? Shmulik. -Original Message- From: Jeff Garzik [mailto:[EMAIL PROTECTED]] Sent: Thursday, November 09, 2000 7:37 PM To: Hen, Shmulik Cc: 'LNML'; 'LKML'; [EMAIL PROTECTED] Subject: Re: catch 22 - porting net driver from 2.2 to 2.4 do_ioctl is inside rtnl_lock... Remember if you need to alter the rules, you can always queue work in the current context, and have a kernel thread handle the work. The nice thing about a kernel thread is that you start with a [almost] clean state, when it comes to locks. Jeff -- Jeff Garzik | Building 1024 | Would you like a Twinkie? MandrakeSoft| - To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to [EMAIL PROTECTED] - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: catch 22 - porting net driver from 2.2 to 2.4
Where can I find info about that ? My first idea was to fire a timer and let the callback routine do the work, but I worry about synchronization and about passing the list of items for it to handle. What is the accepted way of starting a kernel thread and how do I handle parameters and sync. ? Thanks, Shmulik. -Original Message- From: Jeff Garzik [mailto:[EMAIL PROTECTED]] Sent: Thursday, November 09, 2000 7:37 PM To: Hen, Shmulik Cc: 'LNML'; 'LKML'; [EMAIL PROTECTED] Subject: Re: catch 22 - porting net driver from 2.2 to 2.4 do_ioctl is inside rtnl_lock... Remember if you need to alter the rules, you can always queue work in the current context, and have a kernel thread handle the work. The nice thing about a kernel thread is that you start with a [almost] clean state, when it comes to locks. Jeff -- Jeff Garzik | Building 1024 | Would you like a Twinkie? MandrakeSoft| - To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to [EMAIL PROTECTED] - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: catch 22 - porting net driver from 2.2 to 2.4
"Jeff Garzik" wrote: > Theoretically, if you call unregister_netdev from rmmon, it should grab > rtnl_lock and then complete the operation for you. If that doesn't work > for you, it sounds like you are not setting up, or cleaning up, > something correctly. > > Basically... it sounds like there are still bugs in your driver that > need working out :) I followed the value of dev->refcnt and there is something strange. before the call to register_netdev it is set to 0 and after that it is increased to 1. but before the call to unregister_netdev it is somehow 2. How can I tell who is modifying it and when ? I tried using kdb to find out but it keeps hanging the machine. I wanted to place a breakpoint that will pop if a certain memory address is accessed so I did the following: static void *p; (global - at the top of my source file) EXPORT_SYMBOL (p); in my probe function: p = (void*) dev->refcnt; > insmod my_module > ksyms this showed that p is at address 0xd081ee90 then from within kdb: kdb> md 0xd081ee90 0xd081ee90 c470bedc kdb> bpha 0xc470bedc Forced Global Breakpoint at... kdb> go system hung Is there a way to place a breakpoint on a memory address access ? - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Q: using kdb to trap memory modifications
hello, I'm trying to see when and how a certain variable is being modified and I wonder how to get kdb to do that for me. when I do 'insmod my_module' with my network driver, I noticed that dev->refcnt is 0 at first and gets increased to 1 after calling register_netdev. When I want to do 'rmmod my_module' somehow dev->refcnt is at 2 before the call to unregister_netdev and so the kernel (2.4.0-test9) won't let the module to unload. Since I don't increase it explicitly, I want to know who does it for me and when to see if I'm doing anything wrong. So, I modified my module to contain in the global section: static void* ptr; EXPORT_SYMBOL(ptr); and in my module_init function I added: ptr = (void*) >refcnt; after running 'insmod my_module' I used ksyms and found that ptr is at address 0xd081eeb4. after entering kdb, I did: kdb> md 0xd081eeb4 0xd081eeb4 c470bedc 0xd081eec4 . . kdb> bph 0xc470bedc Forced Breakpoint at #0 ... kdb> be 0 kdb> go When I try to return to normal shell, the system is totally hung and won't even receive inputs from the remote serial terminal. I tried bp, bpa and bpha, but the result is always the same. Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: catch 22 - porting net driver from 2.2 to 2.4
I tried using the kernel thread as demonstrated in your example and again it failed (panic - scheduling in interrupt). The difference is that your code executes the thread from within dev->open, while my code tries to do that from dev->do_ioctl that has spinlocks around the entire operation (which apparently sleeps). If I comment out the spin_lock/unlock it will succeed, but then I can't be sure I don't get any concurrent Tx/Rx/timer which is a bad idea while the topology is still being created. is there any way to do something like firing threads/timers atomically ? Thanks, Shmulik. -Original Message- From: Jeff Garzik [mailto:[EMAIL PROTECTED]] Sent: Monday, November 13, 2000 4:26 PM To: Hen, Shmulik Subject: Re: catch 22 - porting net driver from 2.2 to 2.4 "Hen, Shmulik" wrote: > > Where can I find info about that ? > My first idea was to fire a timer and let the callback routine do the work, > but I worry about synchronization and about passing the list of items for it > to handle. > What is the accepted way of starting a kernel thread and how do I handle > parameters and sync. ? Attached is an example. My "8139too" ethernet driver uses a kernel thread instead of a timer to perform media checking. It illustrates how to start and stop a kernel thread. -- Jeff Garzik | Building 1024 | The chief enemy of creativity is "good" sense MandrakeSoft| -- Picasso - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
kernel memory allocations alignment
Hello, When using kmalloc(size_t size), do I get a guaranty that the memory region allocated is aligned according to the size specified ? More to the point, if I call kmalloc for type int on an IA64 architecture is the pointer going to be 8 bytes aligned ? Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
RE: kernel memory allocations alignment
Actually yes. We were warned that on IA64 architecture the system will halt when accessing any type of variable via a pointer if the pointer does not contain an aligned address matching that type. Until now we were using a method of receiving a pointer to an array, casting it to a pointer of a struct (packed with #pragma pack(1) ) ,and retrieving fields directly from it with pointers. It seems we cannot do that any more and were wondering what are the alternatives. One way we could think of is forget the packing and rearrange the fields in the struct in descending order so they all come out aligned, but we didn't know for sure if the first one will be aligned too. Will that work ? Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Intel Network Communications Group Jerusalem, Israel -Original Message- From: Manfred [mailto:[EMAIL PROTECTED]] Sent: Sunday, February 04, 2001 5:56 PM To: Hen, Shmulik Cc: 'LKML' Subject: Re: kernel memory allocations alignment "Hen, Shmulik" wrote: > > When using kmalloc(size_t size), do I get a guaranty that the memory region > allocated is aligned according to the size specified ? > More to the point, if I call kmalloc for type int on an IA64 architecture is > the pointer going to be 8 bytes aligned ? > Yes, kmalloc results are always 'sizeof(void*)' aligned. Do you have stricter alignment requirements? -- Manfred - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] Please read the FAQ at http://www.tux.org/lkml/
Q: How do I get from the latest stable kernel version to the latest prepatch version ?
Hi, According to http://www.kernel.org, the latest stable kernel version is 2.4.2. The latest prepatch version is 2.4.3-pre3. In order to get a full 2.4.3-pre8 kernel do I have to: A. download linux-2.4.2.tar.gz and all the patch-2.4.3-preX.gz and apply them in succession or, B. download linux-2.4.3.tar.gz (exists ?) and then apply the all patches or, C. download linux-2.4.3-pre7.tar.gz (exists ?) and apply only patch-2.4.3-pre8.gz ? Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) Intel Corporation Ltd. - 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/
RE: Q: How do I get from the latest stable kernel version to the latest prepatch version ?
Thanks. It just struck me odd that the latest is 2.4.2 while the prepatches were 2.4.3 so I figured there must be something I missed in between (my logic told me that a 2.4.3 patch would be against a 2.4.3 something ;-). BTW, I haven't seen any announcements from Linus in this mailing list regarding new versions, just the updates on the web site and Alan's release notes saying he's merging with 2.4.3xx. Are those announcements being posted somewhere else now ? -Original Message- From: Leonid Mamtchenkov [mailto:[EMAIL PROTECTED]] Sent: Monday, March 26, 2001 2:33 PM To: Hen, Shmulik Cc: 'LKML' Subject: Re: Q: How do I get from the latest stable kernel version to the late st prepatch version ? Hello Hen, Shmulik, Once you wrote about "Q: How do I get from the latest stable kernel version to the late st prepatch version ?": HS> According to http://www.kernel.org, the latest stable kernel version is HS> 2.4.2. The latest prepatch version is 2.4.3-pre3. HS> HS> In order to get a full 2.4.3-pre8 kernel do I have to: HS> HS> A. download linux-2.4.2.tar.gz and all the patch-2.4.3-preX.gz and apply HS> them in succession or, HS> B. download linux-2.4.3.tar.gz (exists ?) and then apply the all patches or, HS> C. download linux-2.4.3-pre7.tar.gz (exists ?) and apply only HS> patch-2.4.3-pre8.gz ? Download 2.4.2 and then apply 2.4.3-preX (latest) on it... that's it. You might want to visit http://kernelnewbies.org . They have some good docs there. -- Best regards, Leonid Mamtchenkov System Administrator - 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/
RE: URGENT : System hands on "Freeing unused kernel memory: "
Does it hang forever ? I've noticed that my kernel (2.4.2) stalls for several minutes with the same message but suddenly after that the login prompt appears (anything between, like configurations and services starting messages, are gone). We've been able to track it down to a change we did to /etc/lilo.conf to add support for kernel prints to go out to a serial debugger. Before that everything was OK, but after we added append="console=tty0 console=ttyS1,38400", this problem started. We did notice however that everything that doesn't appear on the console does appear on the serial debugger. Shmulik. -Original Message- From: Thomas Foerster [mailto:[EMAIL PROTECTED]] Sent: Tuesday, March 27, 2001 10:40 AM To: [EMAIL PROTECTED] Subject: Re: URGENT : System hands on "Freeing unused kernel memory: " > On 03.27 Thomas Foerster wrote: >> >> But suddenly the box was offline. One technical assistant from our ISP tried >> to reboot >> our server (he couldn't tell me if there had been any messages on the screen), >> but the >> system always hangs on >> >> Freeing unused kernel memory: xxk freed >> > Try booting with init=/bin/bash, it looks like kernel gets a bad /sbin/init, > and gets stuck. Perhaps the shutdown damaged init, it starts to run and get > hung. That didn't fix the problem :( When i run "diff" on a new and the "old" init, i get no diffs ... Must be something other :( Thomas - 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/ - 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/
RE: Plans for 2.5
Just some general questions: 1) Is there anywhere a list that describes what is intended to be in 2.5.x ? 2) Are there any early releases of 2.5.x ? 3) Are the things for 2.5.x being discussed on another mailing list ? 4) What is the time frame of releasing 2.5.x-final (or 2.6.x) ? Specifically, I'm more interested in the network driver aspect. 1) Are there any intended changes to the networking layer ? 2) I over heard something about making the driver reentrant - any news ? 3) What about support for IPv6 ? (I noticed it was marked as experimental until now) Thanks in advance, Shmulik Hen Software Engineer Linux Advanced Networking Services Intel Network Communications Group Jerusalem, Israel -Original Message- From: Bruno Avila [mailto:[EMAIL PROTECTED]] Sent: Thursday, March 29, 2001 12:45 AM To: [EMAIL PROTECTED] Subject: Plans for 2.5 Hello people, I got some questions. When are we going to develop stuff for 2.5? What is planed? My opinion for linux 2.5 should be performance. Since linux already is stable or well done for nature, we could thing more on performance to be a diferencial over others. What do you people thing? Bruno Avila PS: Not a good english. I know! :) - 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/ - 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/
change_mtu boundary checking error
Hello, Going through the change_mtu() code in the kernel, I came across the default function supplied when calling ether_setup(). I could see that eth_change_mtu() (drivers/net/net_init.c) does the following: if( (new_mtu < 68) || (new_mtu > 1500) ) return -EINVAL; Looking in include/linux/if_ether.h I found the following constants: #define ETH_ALEN6 /* Octets in one ethernet addr */ #define ETH_HLAN14 /* Total octets in header. */ #define ETH_ZLEN60 /* Min. octets in frame sans FCS */ #define ETH_DATA_LEN1500/* Max. octets in payload */ #define ETH_FRAME_LEN 1514/* Max. octets in frame sans FCS */ Now, the high boundary seemed reasonable (ETH_FRAME_LEN - ETH_HLEN = ETH_DATA_LEN) which gives 1500, but why is the low boundary set to 68 ? According to my calculations, it should have been ETH_ZLEN - ETH_HLEN which gives 46. Doesn't mtu means only the payload size ? Where did the 68 come from ? Thanks, Shmulik Hen Software Engineer Linux Advanced Networking Services Network Communications Group, Israel (NCGj) - 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/
RE: change_mtu boundary checking error
But Ethernet is not only for IP, what about other protocols ? -Original Message- From: Alan Cox [mailto:[EMAIL PROTECTED]] Sent: Tuesday, April 17, 2001 3:41 PM To: [EMAIL PROTECTED] Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: Re: change_mtu boundary checking error > Now, the high boundary seemed reasonable (ETH_FRAME_LEN - ETH_HLEN = > ETH_DATA_LEN) which gives 1500, but why is the low boundary set to 68 ? > According to my calculations, it should have been ETH_ZLEN - ETH_HLEN which > gives 46. The IPv4 minimum MTU is 68 bytes. Below that not all frames can be delivered - 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/
