April 2005 - Darwin-kernel - lists.apple.com

Re: Kernel stack size
by Nikita Danilov 13 Apr '05

13 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Carl Smith writes: > > OK let me try it another way. > > I have this structure declared as global: > Struct Frame_Data > { > void *buff1; > void *buff2; > void *buff3; > } > > Then I declare as a global: > > Int g_Array_Cnt = 0; > > struct Frame_Data my_data[50]; > > So in my kernel's init routine I do: > > While(g_Array_Cnt < 50) > { > if((my_data[g_Array_Cnt].buff1 = _MALLOC(1600, M_TEMP, > M_WAITOK)) != 0) > { > bzero(&my_data[g_Array_Cnt].buff1, 1600); > > if((my_data[g_Array_Cnt].buff2 = _MALLOC(1600, M_TEMP, > M_WAITOK)) != 0) > { > bzero(&my_data[g_Array_Cnt].buff2, 1600); > > if((my_data[g_Array_Cnt].buff3 = _MALLOC(1600, > M_TEMP, M_WAITOK)) != 0) > { > bzero(&my_data[g_Array_Cnt].buff3, > 1600); > } > else > // error handling > } > else > // error handling > } > else > // error handling This (after correcting obvious error pointed to by others in this thread) can be simplified to enum { FRAME_BUF_SIZE = 1600 }; my_data[g_Array_Cnt].buff1 = _MALLOC(FRAME_BUF_SIZE, M_TEMP, M_WAITOK | M_ZERO); my_data[g_Array_Cnt].buff2 = _MALLOC(FRAME_BUF_SIZE, M_TEMP, M_WAITOK | M_ZERO); my_data[g_Array_Cnt].buff3 = _MALLOC(FRAME_BUF_SIZE, M_TEMP, M_WAITOK | M_ZERO); M_ZERO automatically zeroes allocated space, and M_WAITOK is guaranteed to return non-NULL (or at least a _lot_ of places in core XNU code assumes it is.). Identical calls to _MALLOC can be further factored into special frame_alloc_buf() function. Nikita. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

RE: Kernel stack size
by William Kucharski 12 Apr '05

12 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Tuesday, April 12, 2005, at 04:45PM, Carl Smith <CSmith(a)fortresstech.com> wrote: > >OK let me try it another way. > >I have this structure declared as global: >Struct Frame_Data >{ > void *buff1; > void *buff2; > void *buff3; >} > >Then I declare as a global: > >Int g_Array_Cnt = 0; > >struct Frame_Data my_data[50]; > >So in my kernel's init routine I do: > >While(g_Array_Cnt < 50) >{ > if((my_data[g_Array_Cnt].buff1 = _MALLOC(1600, M_TEMP, >M_WAITOK)) != 0) > { > bzero(&my_data[g_Array_Cnt].buff1, 1600); > > if((my_data[g_Array_Cnt].buff2 = _MALLOC(1600, M_TEMP, >M_WAITOK)) != 0) > { > bzero(&my_data[g_Array_Cnt].buff2, 1600); > > if((my_data[g_Array_Cnt].buff3 = _MALLOC(1600, >M_TEMP, M_WAITOK)) != 0) > { > bzero(&my_data[g_Array_Cnt].buff3, >1600); > } > else > // error handling > } > else > // error handling > } > else > // error handling > > g_Array_Cnt++; >} I'm frankly amazed this takes as long as it does to start behaving strangely. You want to be doing: bzero(my_data[g_Array_cnt].buff2); Notice, NO AMPERSAND. Currently you're wiping out 1600 bytes of data space every time you bzero something, not clearing the buffer you just allocated... William Kucharski kucharsk(a)mac.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

RE: Kernel stack size
by Jamie Wood 12 Apr '05

12 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Looks like this might be your problem: Hope this helps, Jamie _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… bzero(&my_data[g_Array_Cnt].buff1, 1600); You are zero-ing out 1600 bytes starting with the address of my_data[g_Array_Cnt].buff1, not the address pointed to by my_data[g_Array_Cnt].buff1. This email sent to site_archiver(a)lists.apple.com

1 0

RE: Kernel stack size
by Carl Smith 12 Apr '05

12 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcU/romlAGr7BSS2TJiw3ycRsWiMQAAAHeCw Thread-topic: Kernel stack size OK let me try it another way. I have this structure declared as global: Struct Frame_Data { void *buff1; void *buff2; void *buff3; } Then I declare as a global: Int g_Array_Cnt = 0; struct Frame_Data my_data[50]; So in my kernel's init routine I do: While(g_Array_Cnt < 50) { if((my_data[g_Array_Cnt].buff1 = _MALLOC(1600, M_TEMP, M_WAITOK)) != 0) { bzero(&my_data[g_Array_Cnt].buff1, 1600); if((my_data[g_Array_Cnt].buff2 = _MALLOC(1600, M_TEMP, M_WAITOK)) != 0) { bzero(&my_data[g_Array_Cnt].buff2, 1600); if((my_data[g_Array_Cnt].buff3 = _MALLOC(1600, M_TEMP, M_WAITOK)) != 0) { bzero(&my_data[g_Array_Cnt].buff3, 1600); } else // error handling } else // error handling } else // error handling g_Array_Cnt++; } As I was saying g_Array_Cnt will go up to 10, go through the first _MALLOC for buff1, then in poof, g_Array_Cnt becomes 0 and the process will keep looping, from 0 to 10. Does that make it clearer? Carl -----Original Message----- From: Justin Walker [mailto:justin@mac.com] Sent: Tuesday, April 12, 2005 6:26 PM To: Carl Smith Subject: Re: Kernel stack size And, right away, I take it off-list :-} On Apr 12, 2005, at 15:09, Carl Smith wrote: > To be more precise I had a global array with three pointers in it to > char* buffers. i.e char* buf1, char* buf2 and so on. > > I set the global array at a size of 50. So I had a 50 structures > contain my three char pointers. I'm having trouble reading the above. It's not quite english :-}. Maybe you need to put a governor on your brain, so it doesn't outrun your fingers :-}. Cheers, Justin -- Justin C. Walker, Curmudgeon-At-Large Institute for General Semantics -------- If you're not confused, You're not paying attention -------- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

FW: Kernel stack size
by Carl Smith 12 Apr '05

12 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcU/qDQv4xQ/IGaoRhqFwEkV6QdqVQAAMp8AAADNP7A= Thread-topic: Kernel stack size Ah thanks Brian, and Justin. To be more precise I had a global array with three pointers in it to char* buffers. i.e char* buf1, char* buf2 and so on. I set the global array at a size of 50. So I had a 50 structures contain my three char pointers. At run time I was creating the buffers dynamically for the char pointers at 1600 bytes each. So that was 4800 bytes I was allocating per loop, when I reached the loop count of 10, I could create the first char buffer in the array, but as I went to create the second char buffer in array 10, I noticed the loop counter, which was a global int, was reset to zero. And where I was getting the kernel panics was when I went to unload the NKE, because I was trying to free the allocated memory but the global counter, g_ArrayIndex, was out of whack. I'll keep looking at the pointer doing after doing the _MALLOC and see if anything is going wrong there, but could my array be too large? I would not think so because I am only declaring pointers, no large amounts of data. Thanks Carl -----Original Message----- From: Brian Bergstrand Sent: Tuesday, April 12, 2005 5:40 PM To: Carl Smith Cc: darwin-kernel(a)lists.apple.com Subject: Re: Kernel stack size -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The kernel stack size is fixed at 16KB (last I heard) per thread. MALLOC does not use stack mem, and incrementing an existing var does not use the stack, and finally if you really were blowing the stack it's an almost certainty that a panic would closely follow. So something else is probably going on. You may be scribbling on the stack via a bad ptr and thus overwriting your counter. On Apr 12, 2005, at 4:31 PM, Carl Smith wrote: > Is there a way to determine the kernel stack size dynamically? > > I am allocating some buffers using _MALLOC and after about a loop of > 10, the global incremental counter resets itself to 0, so I am > thinking I am blowing the stack size. > > Thanks > Carl > Brian Bergstrand <http://www.bergstrand.org/brian/>, AIM: triryche206 PGP Key: <http://www.bergstrand.org/brian/misc/public_key.txt> You cannot make a man by standing a sheep on its hind legs. But by standing a flock of sheep in that position you can make a crowd of men. - - Max Beerbohm As of 04:33:18 PM, iTunes is playing "Once In A Lifetime / The Killing Words" from "Identity Crisis" by "Section 16" -----BEGIN PGP SIGNATURE----- Version: PGP 8.1 iQA/AwUBQlwyI3nR2Fu2x7aiEQISXQCfUnAhImVl+DSr5uxlDRF/6z4x/L8An3ZB JXRrOqWPD2Yh4auVxxgJONMT =5rh0 -----END PGP SIGNATURE----- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Fwd: Kernel stack size
by Brian Bergstrand 12 Apr '05

12 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The kernel stack size is fixed at 16KB (last I heard) per thread. MALLOC does not use stack mem, and incrementing an existing var does not use the stack, and finally if you really were blowing the stack it's an almost certainty that a panic would closely follow. So something else is probably going on. You may be scribbling on the stack via a bad ptr and thus overwriting your counter. On Apr 12, 2005, at 4:31 PM, Carl Smith wrote: > Is there a way to determine the kernel stack size dynamically? > > I am allocating some buffers using _MALLOC and after about a loop of > 10, the global incremental counter resets itself to 0, so I am > thinking I am blowing the stack size. > > Thanks > Carl > Brian Bergstrand <http://www.bergstrand.org/brian/>, AIM: triryche206 PGP Key: <http://www.bergstrand.org/brian/misc/public_key.txt> You cannot make a man by standing a sheep on its hind legs. But by standing a flock of sheep in that position you can make a crowd of men. - - - Max Beerbohm As of 04:33:18 PM, iTunes is playing "Once In A Lifetime / The Killing Words" from "Identity Crisis" by "Section 16" -----BEGIN PGP SIGNATURE----- Version: PGP 8.1 iQA/AwUBQlw0RnnR2Fu2x7aiEQLZpACeKdkJE11cDD9tNQmeThFhM6LbkrQAn1AI fLJ259r6dK2gu05cgQyaPqel =S0Nk -----END PGP SIGNATURE----- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Kernel stack size
by Justin Walker 12 Apr '05

12 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Apr 12, 2005, at 14:31, Carl Smith wrote: Is there a way to determine the kernel stack size dynamically? If you blow the stack size, you will know, since the kernel will panic. Cheers, Justin -- Justin C. Walker, Curmudgeon-At-Large Institute for General Semantics -------- If you're not confused, You're not paying attention -------- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… I am allocating some buffers using _MALLOC and after about a loop of 10, the global incremental counter resets itself to 0, so I am thinking I am blowing the stack size. I'm not sure I understand. How does using _MALLOC affect the stack? Also, what is "the global incremental counter"? In any case, the kernel stack size is *small*, but as long as you don't declare large arrays on the stack, or go deeply recursive, you should be fine. This email sent to site_archiver(a)lists.apple.com

1 0

Kernel stack size
by Carl Smith 12 Apr '05

12 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcU/pwf9GdpiqgtoTUaMtbjeptfNdA== Thread-topic: Kernel stack size _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Interface filtering
by Justin Walker 12 Apr '05

12 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Apr 12, 2005, at 6:26, Carl Smith wrote: a specific interface are either all from a wireless connection, or none are. Perhaps I am missing the point of your question." Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large Institute for General Semantics -------- It's not whether you win or lose... It's whether *I* win or lose. -------- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… " Since you are using an interface filter, then the packets you get from Well you could have a situation where the machine is both wired, DSL cable or something, and unwired, having a wireless nic card installed, so in my case there is a en0 and a en1. So if I only wanted to intercept the packets from the wireless card, I was wondering if there was some way I could do that. But it sounds like maybe not. Each packet will come from its originating interface, and you can distinguish that way. You will have to dig further to find out how to identify wireless interfaces. I'm sure that, if the driver knows and publishes the fact, you can find it out (I just don't know the way to do that). This email sent to site_archiver(a)lists.apple.com

1 0

RE: Interface filtering
by Carl Smith 12 Apr '05

12 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcU+8R0Q8pHoKVbqS1GvkAvCaK5GoAAcNrqw Thread-topic: Interface filtering " Since you are using an interface filter, then the packets you get from a specific interface are either all from a wireless connection, or none are. Perhaps I am missing the point of your question." Well you could have a situation where the machine is both wired, DSL cable or something, and unwired, having a wireless nic card installed, so in my case there is a en0 and a en1. So if I only wanted to intercept the packets from the wireless card, I was wondering if there was some way I could do that. But it sounds like maybe not. Thanks for the information Carl -----Original Message----- From: darwin-kernel-bounces+csmith=fortresstech.com(a)lists.apple.com [mailto:darwin-kernel-bounces+csmith=fortresstech.com@lists.apple.com] On Behalf Of Justin Walker Sent: Monday, April 11, 2005 7:50 PM To: darwin-kernel Dev Subject: Re: Interface filtering On Apr 11, 2005, at 16:23, Carl Smith wrote: > I have my Ethernet interface filtering in place and am trapping the > packets as they come through the interface filter, I was just wondering > if there was some way to tell which packets from the interface may be > from a wireless connection or not? Since you are using an interface filter, then the packets you get from a specific interface are either all from a wireless connection, or none are. Perhaps I am missing the point of your question. > In my filter I am passed a struct ifnet and the mbuf. I see that in the > struct ifnet it gives me the interface name, i.e. 'en', 'lo' and so on, > and with this I was just wondering if there is some variable in the > struct that might also tell me if it is wireless or not. There is nothing in the name that will give you a hint, and all will identified as ethernet, at the "BSD network device" layer. There is probably a mechanism available to distinguish wireless devices from others, at the IOKit layer, but I don't know what it is (perhaps the archives will tell). There is provision for maintaining this info at the BSD/network layer (the if_data substructure of the ifnet structure), but I don't know that this is ever filled in, and in any case, this would be done by the driver, so I can't say you can count on it. Anyone else? Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large Institute for General Semantics -------- "Weaseling out of things is what separates us from the animals. Well, except the weasel." - Homer J Simpson -------- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/csmith%40fortresste ch.com This email sent to csmith(a)fortresstech.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0