Re: wi0 and mtu setting [bad idea]
On Fri, 3 Jan 2003 21:28:35 +0200 (WET) Evren Yurtesen wrote: Isnt it also the responsibility of the person who sets the MTU that he should be sure everything will work right? in my access points setting MTU to higher than 1500 works for example. I am using linux based access points. For these last 20 years or so, I have been an IP Network engineer. I assure you that it ends up not being the responsibility of the end user, it is the responsibility of the network staff at the few dozen ISPs a given connection traverses. Bad MTUs combined with broken MTU detection leads to mysterious failures. The Internet has few governing laws. It is an extraordinary example of international cooperation on an unprecidented scale. There is no law stopping you from using an illegal MTU setting, just convention and engineering good manners. I posted the links to the wi (802.llb) standards. In a brief scan of the document, I did not see anything in there allowing MTUs greater than 1500 octets. In the networking world, it is considered very bad to justify a configuration/feature because a given implementation allows the configuration. This leads to networks that don't work. If the wi standard requires interfaces to allow MTUs greater than 1500 octets, and the FreeBSD wi driver doesn't allow them, then wi is broken. If the wi standard optionally allows MTUs 1500 octets, then the wi driver may be uncompetitive. If the wi standard doesn't allow MTUs 1500 octets, and the Linux driver does, then the Linux driver is majorly broken. So, I don't know the answer to whether MTUs greater than 1500 octets are legal under the standard. The standard is the only valid source of information. Supply this information and we can move forward. regards, fletcher To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: wi0 and mtu setting [bad idea]
I think something is called broken when it doesnt work as you expect it to work or it doesnt comply with standards. Also I think something is called an 'extra feature' if it is not in the standard which defines it but you have access to it. Can you please show me the 802.11b document where it says the MTU shouldnt be more than 1500bytes? Maybe it doesnt say that the MTU can be more than 1500 bytes but I think the standards also doesnt say if it must be smaller than 1500 bytes. Also by lettin user to be able to set MTU higher than 1500 you do not break the wi driver. You give an extra feature to it. It is the user who breaks the standards(if it says in 802.11b standards that the MTU cant be bigger than 1500) if he/she sets MTU higher than 1500 Evren On Mon, 6 Jan 2003 [EMAIL PROTECTED] wrote: On Fri, 3 Jan 2003 21:28:35 +0200 (WET) Evren Yurtesen wrote: Isnt it also the responsibility of the person who sets the MTU that he should be sure everything will work right? in my access points setting MTU to higher than 1500 works for example. I am using linux based access points. For these last 20 years or so, I have been an IP Network engineer. I assure you that it ends up not being the responsibility of the end user, it is the responsibility of the network staff at the few dozen ISPs a given connection traverses. Bad MTUs combined with broken MTU detection leads to mysterious failures. The Internet has few governing laws. It is an extraordinary example of international cooperation on an unprecidented scale. There is no law stopping you from using an illegal MTU setting, just convention and engineering good manners. I posted the links to the wi (802.llb) standards. In a brief scan of the document, I did not see anything in there allowing MTUs greater than 1500 octets. In the networking world, it is considered very bad to justify a configuration/feature because a given implementation allows the configuration. This leads to networks that don't work. If the wi standard requires interfaces to allow MTUs greater than 1500 octets, and the FreeBSD wi driver doesn't allow them, then wi is broken. If the wi standard optionally allows MTUs 1500 octets, then the wi driver may be uncompetitive. If the wi standard doesn't allow MTUs 1500 octets, and the Linux driver does, then the Linux driver is majorly broken. So, I don't know the answer to whether MTUs greater than 1500 octets are legal under the standard. The standard is the only valid source of information. Supply this information and we can move forward. regards, fletcher To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: wi0 and mtu setting [bad idea]
In message: [EMAIL PROTECTED] Evren Yurtesen [EMAIL PROTECTED] writes: : Yes, actually this is why I asked first about MTU settings in wi. I use : PPPoE and it has 8bytes overhead. It could be nice to set MTU to 1508 so : 1500 byte ethernet frames can be encapsualted in PPPoE without : fragmentation. : : I just dont understand why FreeBSD people have to make this wi driver so : tight in standarts even though there are wireless cards which support : non-standard(and sometimes nice) things which can be useful. I definetely : disagree that something should be in standard so you will allow setting of : it! Dude, you are being way too confrontational here. Since I am the wi driver maintainer, and I haven't said I wouldn't support this, you are just being an ass rather than doing things in a way that would motivate people to allow you the flexibility to do. I'm not conversant on all the technical issues, but so far I've missed patches that I could test to see if they cause problems or not. Warner To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: wi0 and mtu setting [bad idea]
In message: [EMAIL PROTECTED] Evren Yurtesen [EMAIL PROTECTED] writes: : I think something is called broken when it doesnt work as you expect it to : work or it doesnt comply with standards. : Also I think something is called an 'extra feature' if it is not in the : standard which defines it but you have access to it. : : Can you please show me the 802.11b document where it says the MTU shouldnt : be more than 1500bytes? Maybe it doesnt say that the MTU can be more than : 1500 bytes but I think the standards also doesnt say if it must be smaller : than 1500 bytes. : : Also by lettin user to be able to set MTU higher than 1500 you do not : break the wi driver. You give an extra feature to it. It is the user who : breaks the standards(if it says in 802.11b standards that the MTU cant be : bigger than 1500) if he/she sets MTU higher than 1500 Actually, if you want to play the standards card, you need to show where in the doc it is allowed. The burdon of proof is on you, not the me, for this. I've seen some compelling arguments to do this, but also a lot of stupid grousing and complaining that make me ill inclined to help. Warner To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: wi0 and mtu setting [bad idea]
Sorry about that but everybody seemed to disagree to having this option of 1500+ MTU settings in wi driver. If you will support this, then its just fine for me. I dont complain! =) Sorry! Evren On Mon, 6 Jan 2003, M. Warner Losh wrote: In message: [EMAIL PROTECTED] Evren Yurtesen [EMAIL PROTECTED] writes: : Yes, actually this is why I asked first about MTU settings in wi. I use : PPPoE and it has 8bytes overhead. It could be nice to set MTU to 1508 so : 1500 byte ethernet frames can be encapsualted in PPPoE without : fragmentation. : : I just dont understand why FreeBSD people have to make this wi driver so : tight in standarts even though there are wireless cards which support : non-standard(and sometimes nice) things which can be useful. I definetely : disagree that something should be in standard so you will allow setting of : it! Dude, you are being way too confrontational here. Since I am the wi driver maintainer, and I haven't said I wouldn't support this, you are just being an ass rather than doing things in a way that would motivate people to allow you the flexibility to do. I'm not conversant on all the technical issues, but so far I've missed patches that I could test to see if they cause problems or not. Warner To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: wi0 and mtu setting [bad idea]
In message: [EMAIL PROTECTED] Evren Yurtesen [EMAIL PROTECTED] writes: : Sorry about that but everybody seemed to disagree to having this option of : 1500+ MTU settings in wi driver. : : If you will support this, then its just fine for me. I dont complain! =) : Sorry! Well, this converstation needs to take place on the net@ list. In -current, at least, if_ethersubr.c prevents 'ethernet' drivers from setting the mtu higher. There's some experimental patches that make 802.11 its own network thing which were ported over from NetBSD, but even those don't allow one to set it higher since it just uses the if_ethersubr code path for mtu setting. So while I'll make it possible for the wi driver, it won't help at all until this higher level stuff is resolved. Warner To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
Re: wi0 and mtu setting [bad idea]
M. Warner Losh writes: Well, this converstation needs to take place on the net@ list. In -current, at least, if_ethersubr.c prevents 'ethernet' drivers from setting the mtu higher. There's some experimental patches that make It doesn't prevent anything. It just sets the mtu to a default of ETHER_MTU, which is what 98% of ethernet drivers want. If you want something higher (or lower) as your default, you just need to reset the mtu after calling ether_ifattach(). For example, the following code at the bottom of my attach function has worked fine for years..: (GM_IP_MTU is 9K on some firmware, 4K on older firmware) #if (__FreeBSD_version = 50) ether_ifattach (ifp, sc-is_addr); #elif (__FreeBSD_version = 41) ether_ifattach (ifp, ETHER_BPF_SUPPORTED); #else if_attach (ifp); ether_ifattach (ifp); #endif /*ether_ifattach resets mtu */ ifp-if_mtu = GM_IP_MTU; return 0; Drew To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message
RE: wi0 and mtu setting [bad idea]
Yes, actually this is why I asked first about MTU settings in wi. I use PPPoE and it has 8bytes overhead. It could be nice to set MTU to 1508 so 1500 byte ethernet frames can be encapsualted in PPPoE without fragmentation. I just dont understand why FreeBSD people have to make this wi driver so tight in standarts even though there are wireless cards which support non-standard(and sometimes nice) things which can be useful. I definetely disagree that something should be in standard so you will allow setting of it! For example in /etc/defaults/rc.conf it says --- tcp_drop_synfin=NO# Set to YES to drop TCP packets with SYN+FIN # NOTE: this violates the TCP specification --- So why do you let this to violate TCP specifications but you cant let user to violate ethernet specifications? Evren On Sat, 4 Jan 2003, Sten Daniel Sørsdal wrote: How about a configuration of two Ad-hoc cards pointing towards eachother between two buildings and an IPSec tunnel is applied. Wouldn't it be great if (unencrypted) packets destined to go through that IPSec tunnel could go through in full ethernet size, without fragmentation, pr host tcp stack adjustments or resending because of DF flag? What about transporting VLANs over wireless? There is a lot of equipment out there, especially wireless but also wired (ATM?) that allows larger MTUs for special circumstances. It's like buying a car with all the extra features - but only a handful of the features work. Just my 2 nkr --- Med vennlig hilsen / Best regards Sten Daniel Sørsdal Wireless Manager WAN Norway AS --- -Original Message- From: Wright, Michaelx L [mailto:[EMAIL PROTECTED]] Sent: 3. januar 2003 19:28 To: Evren Yurtesen; [EMAIL PROTECTED] Cc: Michael Sierchio; [EMAIL PROTECTED]; [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: RE: wi0 and mtu setting [bad idea] Good Afternoon All, I am curious to know if you are taking into account MTU limitations imposed by link-partners i.e. switches, hubs, routers and the like. Some if not most ( for Unix) require end-nodes to be approximately 22 bytes less than the link-partner device's maximum supported MTU. I am not sure if, but would somewhat expect, a wireless access point to have some impact on the sizing and/transfer at above the 1500 MTU setting. Cheers M. L. Wright Intel UNIX-NQL 503.264.8300 -Original Message- From: Evren Yurtesen [mailto:[EMAIL PROTECTED]] Sent: Friday, January 03, 2003 10:07 AM To: [EMAIL PROTECTED] Cc: Michael Sierchio; [EMAIL PROTECTED]; [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: Re: wi0 and mtu setting [bad idea] You are definetely right, setting the MTU might be really bad thing, but why dont you let the person setting it decide it for himself? Thus FreeBSD wi driver can support setting this value higher than 1500 in your own risk. Its a functionality request only. I dont suggest that you set the default mtu for wi driver something higher than 1500! Evren On Fri, 3 Jan 2003 [EMAIL PROTECTED] wrote: On Fri, 3 Jan 2003 02:22:34 +0200 (WET) Evren Yurtesen wrote: I definetely agree and obviously since mikrotikos supports this then linux should do since mikrotikos is built on linux. Why shouldnt FreeBSD support setting mtu of wireless interfaces higher than 1500 Setting a wireless interface to a MTU of higher than 1500 octets is ill-advised unless you are in very specific, unusual conditions. The subject header talks about wi0, which implies IEEE Ethernet 802.11b standard interface. The IEEE maintains the Ethernet standards. Start with: http://www.ieee.org or http://www.ieee802.org From a quick glance at the standard: IEEE Std 802.11b-1999 (Supplement to ANSI/IEEE Std 802.11, 1999 Edition) Supplement to IEEE Standard for Information technology Telecommunications and information exchange between systems Local and metropolitan area networks Specific requirements Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: Higher-Speed Physical Layer Extension in the 2.4 GHz Band it is not clear to me that MTU 1500 octets are legal with 802.11b. If your system is connected to the Internet, setting the MTU on your FreeBSD system, which is probably not a core router, to anything above 1500 is a stupid idea. If you don't already know this, and don't understand the reasons why, you would be best advised not to mess with the MTU at all. Stick with the default until you gain more experience. You might want to read up on packet fragmentation and MTU discovery for explanations why this is a good idea. good luck, fletcher To Unsubscribe: send mail
RE: wi0 and mtu setting [bad idea]
How about a configuration of two Ad-hoc cards pointing towards eachother between two buildings and an IPSec tunnel is applied. Wouldn't it be great if (unencrypted) packets destined to go through that IPSec tunnel could go through in full ethernet size, without fragmentation, pr host tcp stack adjustments or resending because of DF flag? What about transporting VLANs over wireless? There is a lot of equipment out there, especially wireless but also wired (ATM?) that allows larger MTUs for special circumstances. It's like buying a car with all the extra features - but only a handful of the features work. Just my 2 nkr --- Med vennlig hilsen / Best regards Sten Daniel Sørsdal Wireless Manager WAN Norway AS --- -Original Message- From: Wright, Michaelx L [mailto:[EMAIL PROTECTED]] Sent: 3. januar 2003 19:28 To: Evren Yurtesen; [EMAIL PROTECTED] Cc: Michael Sierchio; [EMAIL PROTECTED]; [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: RE: wi0 and mtu setting [bad idea] Good Afternoon All, I am curious to know if you are taking into account MTU limitations imposed by link-partners i.e. switches, hubs, routers and the like. Some if not most ( for Unix) require end-nodes to be approximately 22 bytes less than the link-partner device's maximum supported MTU. I am not sure if, but would somewhat expect, a wireless access point to have some impact on the sizing and/transfer at above the 1500 MTU setting. Cheers M. L. Wright Intel UNIX-NQL 503.264.8300 -Original Message- From: Evren Yurtesen [mailto:[EMAIL PROTECTED]] Sent: Friday, January 03, 2003 10:07 AM To: [EMAIL PROTECTED] Cc: Michael Sierchio; [EMAIL PROTECTED]; [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: Re: wi0 and mtu setting [bad idea] You are definetely right, setting the MTU might be really bad thing, but why dont you let the person setting it decide it for himself? Thus FreeBSD wi driver can support setting this value higher than 1500 in your own risk. Its a functionality request only. I dont suggest that you set the default mtu for wi driver something higher than 1500! Evren On Fri, 3 Jan 2003 [EMAIL PROTECTED] wrote: On Fri, 3 Jan 2003 02:22:34 +0200 (WET) Evren Yurtesen wrote: I definetely agree and obviously since mikrotikos supports this then linux should do since mikrotikos is built on linux. Why shouldnt FreeBSD support setting mtu of wireless interfaces higher than 1500 Setting a wireless interface to a MTU of higher than 1500 octets is ill-advised unless you are in very specific, unusual conditions. The subject header talks about wi0, which implies IEEE Ethernet 802.11b standard interface. The IEEE maintains the Ethernet standards. Start with: http://www.ieee.org or http://www.ieee802.org From a quick glance at the standard: IEEE Std 802.11b-1999 (Supplement to ANSI/IEEE Std 802.11, 1999 Edition) Supplement to IEEE Standard for Information technology Telecommunications and information exchange between systems Local and metropolitan area networks Specific requirements Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: Higher-Speed Physical Layer Extension in the 2.4 GHz Band it is not clear to me that MTU 1500 octets are legal with 802.11b. If your system is connected to the Internet, setting the MTU on your FreeBSD system, which is probably not a core router, to anything above 1500 is a stupid idea. If you don't already know this, and don't understand the reasons why, you would be best advised not to mess with the MTU at all. Stick with the default until you gain more experience. You might want to read up on packet fragmentation and MTU discovery for explanations why this is a good idea. good luck, fletcher To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-stable in the body of the message