November 2008 - Darwin-dev - lists.apple.com

Re: IOUserClient size limit?
by Duane Murphy 11 Nov '08

11 Nov '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 - --- At Tue, 11 Nov 2008 09:47:29 -0800, Kevin Elliott wrote: > >On Nov 10, 2008, at 5:35 PM, Duane Murphy wrote: >> Thanks Chris! I would have completely missed the prepare() and >> complete() methods. >> >> On a follow up question. My user client APIs map to kernel functions >> that expect to get a buffer that they can write into. >> IOMemoryDescriptors are setup to use readBytes() and writeBytes() >> rather >> than direct memory access (e.g. a pointer). > >I really hope your doing something different than what it sounds like >your describing. Your UserClient should be exporting an API specific >to your needs, not a generic copy of kernel API's. Generic API's make >it easier to subvert your UserClient and force it do things contrary >to it's intended purpose. If your API is generic, much of your >security comes from secretiveness- Security through obscurity at it's >worse. Sorry, I used poor wording. "kernel functions" in this case is the cross- platform kernel library for our product. These are functions specifically designed to work across the user/kernel boundary, but in order to be generic, they simply work with buffers. So, no, I am not exporting generic kernel functions. >> Is there a way of getting "buffer" access? Some way of writing >> directly >> to the underlying memory by address? Is that what map() and >> IOMemoryMap >> is for? > >Map gives you pointer level access to the memory described by the >descriptor. > Thanks, I'll have to dig into it a little more to work out how to use it. ...Duane -----BEGIN PGP SIGNATURE----- Version: PGP Desktop 9.9.0 (Build 397) Charset: ISO-8859-1 wsBVAwUBSRnPWUrg9acQ4r2CAQhd9ggA03AdbL8/a41vcwN2EOd8C0a9q0Ojqj7f CyOR1kNvSJuRcy7XlAuxbQ7LQSNZ1yM1hhxojc0QWO1ju5kPEXvE1hlXM+OyvIUi CSvjwnpvLyHQdx3NLsNEsTnNTbAMxnhWQaNfnUvCmgaM+NN9oPR3Mcpq/qIaXgwH jCQoF7P7UVoAmuGU+j5mdLb5aNRuet0WyWJOVDD+8fJZvNj2TS99G2CbxKVFMpfa ReTbXLf2nOL2IWKEDMV332Km0XhWYKn0OfhUYiWqFZtlAmILXs+GXG+UGpSvTdh0 QIPikWJHSHXz4Vxf1xuV1ayQnPGxipFqZiLLk0DaEOQX+iT0BBAo6w== =SHZq -----END PGP SIGNATURE----- _______________________________________________ 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: IOUserClient size limit?
by Kevin Elliott 11 Nov '08

11 Nov '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thanks Chris! I would have completely missed the prepare() and complete() methods. On Nov 10, 2008, at 5:35 PM, Duane Murphy wrote: On a follow up question. My user client APIs map to kernel functions that expect to get a buffer that they can write into. IOMemoryDescriptors are setup to use readBytes() and writeBytes() rather than direct memory access (e.g. a pointer). I really hope your doing something different than what it sounds like your describing. Your UserClient should be exporting an API specific to your needs, not a generic copy of kernel API's. Generic API's make it easier to subvert your UserClient and force it do things contrary to it's intended purpose. If your API is generic, much of your security comes from secretiveness- Security through obscurity at it's worse. Is there a way of getting "buffer" access? Some way of writing directly to the underlying memory by address? Is that what map() and IOMemoryMap is for? Map gives you pointer level access to the memory described by the descriptor. _______________________________________________ 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: IOUserClient size limit?
by Duane Murphy 11 Nov '08

11 Nov '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com --- At Mon, 10 Nov 2008 16:53:42 -0800, Chris Sarcone wrote: >Duane -- > >I am seeing a kIOReturnIPCError form an IOUserClient down call (i.e. >from my app into the kernel) when I have a structure that is over 4K >(4096) bytes in size. > >Is this a hard limit? Can I change this limit? I have some I/O through >the user client that I would like to take a larger buffer (16K). > >Yes, it's a hard limit for IPC. > >Typically, you solve this by: > >1) Save the task_t passed to your user client in initWithTask(). >2) Pass a pointer to the large buffer and use IOMD::withAddressRange() >(which takes the task_t parameter) >3) prepare() it >4) Use it for DMA in conjunction with an IODMACommand or use readBytes()/ >writeBytes() >5) complete() it >6) release the IOMD Thanks Chris! I would have completely missed the prepare() and complete() methods. On a follow up question. My user client APIs map to kernel functions that expect to get a buffer that they can write into. IOMemoryDescriptors are setup to use readBytes() and writeBytes() rather than direct memory access (e.g. a pointer). Is there a way of getting "buffer" access? Some way of writing directly to the underlying memory by address? Is that what map() and IOMemoryMap is for? Otherwise, it seems I'll have to allocate an additional buffer in the kernel to handle the API then copy back into the IOM. Seems like a waste of time and space. Thanks for the help! ...Duane _______________________________________________ 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: IOUserClient size limit?
by Chris Sarcone 11 Nov '08

11 Nov '08

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

Re: IOUserClient size limit?
by Eli Bach 11 Nov '08

11 Nov '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:message-id:from:to :in-reply-to:content-type:content-transfer-encoding:mime-version :subject:date:references:x-mailer; bh=3vajWvdStB7V6uexe1HD3KyTBOTQ1ip2NNDjePhuvg4=; b=JPP8NzZSFgojAEO9z+BrL/dHSW4X0JM6tFA+QcNLXQDUaqi0ypY/BZ9zrLsrhmq2sB evRFq8POnGAQWJfM0roeD6KQsStU3yVestUyG+Z8D5caJtM6NLEE0R6qRU4db4hgYYsP a+j4JGm/tBtWyawiKNBhk7X9DStxbUhYf2S5I= Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:from:to:in-reply-to:content-type :content-transfer-encoding:mime-version:subject:date:references :x-mailer; b=uRe1lwb15Mmv0EscIo9cguteqdFVmFT4SNeOdmO+nJBPSO3ChjKzA4S4OughLa3sf6 DlBqv8lQNMw3Rvb1dKH5C5DgjSx7Vgk52Y01ux/JNg3jrXZSh/7bQfvs2pnCu6P+k+/6 O16Mt4cKYhayHJeGl9B0B46oxaFamDOSPuZWE= On Nov 10, 2008, at 2:05 PM, Duane Murphy wrote: I am seeing a kIOReturnIPCError form an IOUserClient down call (i.e. from my app into the kernel) when I have a structure that is over 4K (4096) bytes in size. Is this a hard limit? Can I change this limit? I have some I/O through the user client that I would like to take a larger buffer (16K). Thanks for any tips, Eli _______________________________________________ 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… ...Duane Yeah, I believe the recommended path to take for passing larger amounts of data between user-space and the kernel is to use a shared buffer that is memory mapped in both the kernel and user-space. And sorry, but offhand I don't know the API to do this. I believe there is sample code for doing this, and if there isn't, I would then check Apple's open-source Darwin driver's. This email sent to site_archiver(a)lists.apple.com

1 0

IOUserClient size limit?
by Duane Murphy 10 Nov '08

10 Nov '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com I am seeing a kIOReturnIPCError form an IOUserClient down call (i.e. from my app into the kernel) when I have a structure that is over 4K (4096) bytes in size. Is this a hard limit? Can I change this limit? I have some I/O through the user client that I would like to take a larger buffer (16K). Thanks for any tips, ...Duane _______________________________________________ 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

CF
by Benjamin Huntsman 10 Nov '08

10 Nov '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thread-index: AclDbJdX9AbqhW5BTBWZd3m8Vz+QXQ== Thread-topic: CF Does anyone know where to obtain CF-196.4 anymore? I thought it used to be available, but the earliest I can find is CF-299. Thanks in advance! _______________________________________________ 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: Finding the architectures of a binary
by Clark Cox 10 Nov '08

10 Nov '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:message-id:date:from:to :subject:cc:in-reply-to:mime-version:content-type :content-transfer-encoding:content-disposition:references; bh=XPv4qESGMStCzlhz3tK4Ez5ao8tynxqRfCzJzWLTdPY=; b=FrrQrFtGeCQq9ZApb/BIB6SDlHUjg1XdumNWxsBtvl+LTUn+psC5D2Wx1BMqLrjbz2 Ec5ik74H2+aP33r6jaJWRTk+wkNPUpO5LtXRBjcmyl4byhzguiAdvmRF0QIC31ElCpMI 4+mS0s/fHh0kBiDnqjzca8u4Fv1mD4WWkUyUA= Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:to:subject:cc:in-reply-to:mime-version :content-type:content-transfer-encoding:content-disposition :references; b=PmK9ZYAjJa5Ibp4R8s+sdY55j+7x7zmEhP1R5V+NBF6yQFO7SPzIKNkSZqklZgNpWN YTFtvs2rFG0/4DR4NJ5+DJt8rI5p1FXg70Mr4kwn1rYj+QyABcFJv2r59tb1Cr5bnGPp whM2Jr5VmHWek/M35xu7R0fSy481VzbE50TJk= On Mon, Nov 10, 2008 at 10:20 AM, Iceberg-Dev <dev.iceberg(a)gmail.com> wrote: > > On Nov 10, 2008, at 7:06 PM, Clark Cox wrote: > >> On Mon, Nov 10, 2008 at 9:23 AM, Iceberg-Dev <dev.iceberg(a)gmail.com> >> wrote: >>> >>> On Nov 10, 2008, at 4:08 PM, Finlay Dobbie wrote: >>>> >>>> >>>> http://www.opensource.apple.com/darwinsource/10.5.5/cctools-667.3/misc/lipo… >>> >>> Thanks. >>> >>> Is it supoosed to be up-to-date and 64-bit compatible? >>> >>> I'm asking because for instance, this line looks suspicious: >>> >>> (*((unsigned long *)addr) == MH_MAGIC || >> >> What looks suspicious about it? > > The ability to run this code as a 64-bit executable. > > The magic cookie is a 32-bit value. AFAIK, if you build a 64-bit executable > and use unsigned long, it's a 64-bit value. > > With a PowerPC, I'm not sure you can guarantee that the higher 4 bytes will > be equal to 0x00000000 as the magic cookie can be the first 4 bytes of a > file (non fat). With an Intel Mac, the issue would be different, since it > would assume that the 4 bytes after the magic cookie in the binary should be > 0x00000000 instead of the cpu_type. Ahh, I see what you mean. No, it does not appear to be 64-bit clean. It probably wouldn't be too hard to hack up your own version that uses unsigned int (or uint32_t, UInt32, etc.) instead of unsigned long -- Clark S. Cox III clarkcox3(a)gmail.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

Re: Finding the architectures of a binary
by Iceberg-Dev 10 Nov '08

10 Nov '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:in-reply-to:references :mime-version:content-type:message-id:cc:content-transfer-encoding :from:subject:date:to:x-mailer; bh=Q5eLnc3yu3skwzfpEzrMcuIX7+8lxodTpB/2bSS3VQc=; b=Suz49ExXNsGk/9GdjVhUuiLh4xMx2bfiQLfg1N9z8Ud5gUrPahNTaNDszlfvozqpv7 n8qkHSD0m1DX3VcNRoO1tPQBoyR+p+v2JxMHUqYNdEDNacO/md9Yp3VcfiAxkB4qfCyd G4SOn90d4dp8syKoiBtLmFh+YsSLLdoRFtsWg= Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=in-reply-to:references:mime-version:content-type:message-id:cc :content-transfer-encoding:from:subject:date:to:x-mailer; b=iAY/WWdNGxF/6AGebJiuak0wW39ZlfIa0NzVCODr2m7Ea1zFZgTaFNwH0BzACL3dyp vSLZr/kybRDvcqKfc9q54WGVjMQnyXf/O9Kixt1DPXaMLjYWXDHkyAjUrUhOTDIrKFXd lNC4nbObqFEKpEgNdB3Mo3o25x48KmQSWNZjE= On Nov 10, 2008, at 7:06 PM, Clark Cox wrote: Thanks. Is it supoosed to be up-to-date and 64-bit compatible? I'm asking because for instance, this line looks suspicious: The ability to run this code as a 64-bit executable. _______________________________________________ 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… On Mon, Nov 10, 2008 at 9:23 AM, Iceberg-Dev <dev.iceberg(a)gmail.com> wrote: On Nov 10, 2008, at 4:08 PM, Finlay Dobbie wrote: http://www.opensource.apple.com/darwinsource/10.5.5/cctools-667.3/ misc/lipo.c (*((unsigned long *)addr) == MH_MAGIC || What looks suspicious about it? The magic cookie is a 32-bit value. AFAIK, if you build a 64-bit executable and use unsigned long, it's a 64-bit value. With a PowerPC, I'm not sure you can guarantee that the higher 4 bytes will be equal to 0x00000000 as the magic cookie can be the first 4 bytes of a file (non fat). With an Intel Mac, the issue would be different, since it would assume that the 4 bytes after the magic cookie in the binary should be 0x00000000 instead of the cpu_type. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Finding the architectures of a binary
by Clark Cox 10 Nov '08

10 Nov '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:message-id:date:from:to :subject:cc:in-reply-to:mime-version:content-type :content-transfer-encoding:content-disposition:references; bh=kbEzAUdxZAEFXRAWc8nFWeQWF+3yV3G6+9pSkHPCmaQ=; b=xu0xYKKLspueuYQCJnl5Cn0V8wKcVE6Q7AsHdTM987lEtBW+C5UMjIsxYzGzzinHyN Zbd8iyF7GI9sjtmMRlltIG0TBfHel7rwkft1t2M0I3fwHQlR7tpPnP1QX4YiK2iRGxLF smwYxDqEg+0hBujsjzMURagigqN142pgwrYds= Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:to:subject:cc:in-reply-to:mime-version :content-type:content-transfer-encoding:content-disposition :references; b=BOTnlsDPQSZ/+852EK0UZioO8HRGLKkktZeg01njqs/+lC+y9UWqqh8pRRSSWviOYs P3M7MaMLuJJ2k50G8j8u17fZ5edGTkVUbDhbwjA+E0GIETbJHUuss8lL3WG4c2XXXZ2q jLXnfAeBNkPhT5Fjf7PF+IiKuaMteE1kAkKq8= On Mon, Nov 10, 2008 at 5:48 AM, Iceberg-Dev <dev.iceberg(a)gmail.com> wrote: > #if BYTE_ORDER==LITTLE_ENDIAN > tMagicCookie=CFSwapInt32(tMagicCookie); > #endif Instead of these preprocessor checks, why not just do: tMagicCookie =CFSwapInt32BigToHost(tMagicCookie); -- Clark S. Cox III clarkcox3(a)gmail.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