May 2007 - Darwin-dev - lists.apple.com

Re: Memory Allocation in Interrupt Handlers?
by Alexander von Below 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thanks for all of your replies! Alex Am 09.05.2007 um 23:55 schrieb Michael Smith: On May 9, 2007, at 5:03 AM, darwin-dev-request(a)lists.apple.com wrote: the answer to this is probably easy to find in the right place, but is there a memory allocation routine that is interrupt safe? From the documentation for IOMalloc it is clear that it is not, so is there a function that is? Or would I have to allocate all my memory prior to going into interrupt context? _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… First; you should go to great lengths to avoid allocating memory on any I/O path. A couple of good ideas: Second; unless you are using an IOFilterInterruptEventSource, you are not running any code in interrupt context. IOInterruptEventSource handlers run in "regular" thread context. You can (but per the above, should not make a habit of) allocate memory in the latter. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Memory Allocation in Interrupt Handlers?
by Michael Smith 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 9, 2007, at 5:03 AM, darwin-dev-request(a)lists.apple.com wrote: the answer to this is probably easy to find in the right place, but is there a memory allocation routine that is interrupt safe? From the documentation for IOMalloc it is clear that it is not, so is there a function that is? Or would I have to allocate all my memory prior to going into interrupt context? There are two separate things to consider here. = Mike _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… First; you should go to great lengths to avoid allocating memory on any I/O path. A couple of good ideas: - Keep a pre-allocated pool of your work structures; consider subclassing IOCommand and use an IOCommandPool. If the pool is empty when new work arrives, it's OK to block and better to block up front than halfway into your setup when an allocation fails and you've run the kernel out of address space. - Avoid using buffers for your data. You don't want to copy data anyway, so this will tend to improve the performance of your driver as a side effect. Use scatter/gather lists rather. Second; unless you are using an IOFilterInterruptEventSource, you are not running any code in interrupt context. IOInterruptEventSource handlers run in "regular" thread context. You can (but per the above, should not make a habit of) allocate memory in the latter. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Darwin-dev Digest, Vol 4, Issue 96
by Jason Seng 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

[darwin-dev] Logging OSDictionary?
by Alexander von Below 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hello List, Thanks Alex _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I feel like a compelte n00b asking, but is there a convenient way to dump the contents of an OSDictionary to the console/log ? Or, in the case of IORegistryObjects, can I assume to see exactly what I see in the Registry Explorer? This email sent to site_archiver(a)lists.apple.com

1 0

Re: [darwin-dev] Memory Allocation in Interrupt Handlers?
by chas williams - CONTRACTOR 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com In message <16F823A7-F427-45D2-ADAD-9B64D09CB746(a)gmail.com>,Shawn Erickson writ es: >On May 8, 2007, at 7:39 AM, Alexander von Below wrote: >> Or would I have to allocate all my memory prior to going into >> interrupt context? > >You should make all attempts to allocate memory before hand. it depends on how you are handling your interrupts. if you arent "taking interrupts directly" then the interrupt thread is the same as any other thread so you dont need to do any special with regard to memory. if you are are taking interrupt directly, then you need to worry about anything that might sleep (like allocating memory). see developer.apple.com about doing this. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: pci config space
by Andrew Gallatin 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Alexander von Below writes: > If I may ask: Why do you need only the PCI configuration space? A variety of reasons. Mainly to check for things like PCI parity errors in large clusters, where running a gui app like Reggie is extraordinarily cubmersome, and a command-line tool (like lspci) scales much better. Also to provide a way to upgrade firmware in our NICs in the field w/o forcing people to boot into some OS where config space access is trivial (linux, solaris, windows, *bsd, basically anything but OSX) or to move the NIC into a box already running such an OS. > My suggestion would be to write a simple driver and communicate with > it. That will free you from the availability of the CHUD API. The problem with this is that the only thing I can visualize will work only on x86 (io port access). I can't seem to figure out a way to get "random access" to the PCI bus. Eg, I don't see a way specify a bus/slot/function and access its PCI config space, even in the kernel. Hence, leaching from Chud seemed appealing.. :) Drew _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: pci config space
by Alexander von Below 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com If I may ask: Why do you need only the PCI configuration space? Alex Am 09.05.2007 um 19:16 schrieb Andrew Gallatin: I need to be able to read PCI configuration space from userspace in my own application. As near as I can tell, the only way to access PCI config space is via the Chud private framework's chudReadPCIConfig(). Is this correct? If so, has anybody figured out enough of the Chud API so as to be able to use basic functionality from the private Chud frameworks in their own application? Thanks, Drew _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/below%40mac.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… My suggestion would be to write a simple driver and communicate with it. That will free you from the availability of the CHUD API. You can even load and unload the driver on demand -- provided you get authorization. This email sent to below(a)mac.com This email sent to site_archiver(a)lists.apple.com

1 0

pci config space
by Andrew Gallatin 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com I need to be able to read PCI configuration space from userspace in my own application. As near as I can tell, the only way to access PCI config space is via the Chud private framework's chudReadPCIConfig(). Is this correct? If so, has anybody figured out enough of the Chud API so as to be able to use basic functionality from the private Chud frameworks in their own application? Thanks, Drew _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: catching floating point exception on OS X Intel
by Axel Luttgens 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.7.2) Gecko/20040804 Netscape/7.2 On 9/05/07 16:59, Axel Luttgens wrote: [...] >From the Mac_OSX_Numerics.pdf (which still seems to apply): Axel _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… [...] Sorry, I am currently experiencing very long delays when accessing the ADC site from here, so I couldn't find the link while writing the above. But I finallly got it: <http://developer.apple.com/documentation/Performance/Conceptual/Mac_OSX_Num…> This email sent to site_archiver(a)lists.apple.com

1 0

Re: catching floating point exception on OS X Intel
by Axel Luttgens 09 May '07

09 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.7.2) Gecko/20040804 Netscape/7.2 On 9/05/07 1:29, Kristoffer Eriksen wrote: Hi- [...] [...] >From the Mac_OSX_Numerics.pdf (which still seems to apply): Floating-point exceptions are signaled with exception flags. When an application begins, all floating-point exception flags are cleared and the default rounding direction (round to nearest) is in effect. This is the default environment. When an exception occurs, the appropriate exception flag is set, but the application continues normal operation. Floating-point exception flags merely indicate that a particular event has occurred; they do not change the flow of control for the application. An application can examine or set individual exception flags and can save and retrieve the entire environment (rounding direction and exception flags). #include <stdio.h> #include <stdlib.h> #include <math.h> #include <fenv.h> #include <signal.h> #include <sys/types.h> #include <unistd.h> int main(int argc, char *argv[]) { double x; int retval; pid_t pid; pid = getpid(); /* Just to be sure ;-) */ feclearexcept(FE_ALL_EXCEPT); /* This will set the FE_INEXACT flag */ x = 1.0/(double)pid; retval = fetestexcept(FE_ALL_EXCEPT); feclearexcept(FE_ALL_EXCEPT); /* This will set the FE_INVALID flag */ x = sqrt(-x); retval = fetestexcept(FE_ALL_EXCEPT); if (retval != 0) raise(SIGFPE); printf("%d %d %f\n",retval,pid,x); exit(0); } HTH, Axel _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I tried this question on darwin-x86 to no avail, so I thought I would try here. To summarize, I'm using the feraiseexcept() function from fenv.h, which I had hoped would enable floating point exceptions for the requested flags. I'm a scientist, not a systems programmer, so I may just be doing something stupid here. Any ideas? Function feraiseexcept() doesn't raise an exception, it only raises one or more exception flags. So, taking your example, but made somewhat more verbose for seeing what's happening: This email sent to site_archiver(a)lists.apple.com

1 0