IRC channel logs

2023-07-06.log

back to list of logs

<damo22>youpi1: what does it mean to map an IO region? i thought rump maps the AHCI bar as a region, and it does i/o on it
<damo22>wont jlledom's patch break rumpdisk?
<damo22>ah AHCI region is not a set of io ports
<youpi1>yes, it's an mmio
<youpi1>not io
<damo22>yep
<damo22>does non-privileged user have ioperm() access to non-locked ports?
<youpi1>ioperm is root-only
<damo22>how will non-root user use a pci card then
<youpi1>the idea of the rpc proposal is to let pci-arbiter possibly give selective ioperm rights
<damo22>so we still have work to do for this
<damo22>io ports are pretty hairy
<damo22>ie, you dont want random poking around in there
<damo22>and possibly two users cant share io ports
<damo22>the same ones i mean
<youpi1>the kernel part is already there
<youpi1>the mail is just about proposing an rpc to pass one
<damo22>one single io port?
<damo22>per request?
<damo22>sometimes io ports work in pairs, an accessor and a data port
<youpi1>no, one range
<damo22>a hok
<damo22>makes more sense
<damo22>so if the range is requested via the kernel with i386_io_perm_create() does the kernel manage the ranges?
<damo22>so it will reject a second request for an overlapping range?
<youpi1>it manages the ranges, yes
<youpi1>I don't remember whether it refuses conflicts
<damo22>* Request a new port IO_PERM that represents the capability to access
<damo22> the I/O ports [FROM; TO] directly. MASTER_PORT is the master device port.
<damo22>i dont think it refuses conflicts for overlapping ranges except the PCI_CFG range
<damo22>is there an efficient way to manage ranges in binary?
<damo22>i bet there is some bitop magic
<youpi1>the documentation doesn't necessarily reflect all of the semantic
<youpi1>for i/o it's easy since it's 4096 ports, so a bitmap is enough yes
<damo22>65536 ports ?
<damo22>but the range of binary is not the same as the bitmap
<damo22>eg, 0xa to 0xf is b1010 to b1111
<damo22>so 0x9 is outside that range (b1001) but 0xb is inside that range (b1011) , i dont see how the bitmap helps
<p4r4D0xum>!n