How hard would it be to have is-enabled kernel probes? While tracing scsi packets I can get lots of really good data by putting probes in scsi_transport() and scsi_hba_pkt_comp() as they contain the scsi packet. However the scsi_address that is available is opaque to all except the HBA and only available in a meaningful form by calling scsi_device_unit_address(). Clearly doing this when the probe is not enabled would be very expensive so having an is-enabled kernel probe would solve this. --chris -- This message posted from opensolaris.org
Chris Gerhard wrote:> How hard would it be to have is-enabled kernel probes? > > While tracing scsi packets I can get lots of really good data by putting probes in scsi_transport() and scsi_hba_pkt_comp() as they contain the scsi packet. However the scsi_address that is available is opaque to all except the HBA and only available in a meaningful form by calling scsi_device_unit_address(). Clearly doing this when the probe is not enabled would be very expensive so having an is-enabled kernel probe would solve this.I thought that object translations (such as for printing) were supposed to be done in user space -- meaning that instead of an "is-enabled" test in the kernel to avoid "expensive" argument set-ups, you should just unload the raw blob and have a dtrace extension to decode that data. -- James Carlson 42.703N 71.076W <carlsonj at workingcode.com>
Unfortunately once in userland there is not enough information to get a user friendly address. The only thing that can convert from the opaque to a string is the HBA driver where each HBA driver can implement their own method. Since we don''t even own all the drivers I can''t go messing in there hence the desire to use an is-enabled probe --chris -- This message posted from opensolaris.org
Hey Chris,> How hard would it be to have is-enabled kernel probes?It''s something we''ve been talking about for awhile; it''s not simple. If a member of the DTrace team had a month or two, I think we could knock it out, but I don''t anticipate that uninterrupted span happening soon. Adam -- Adam Leventhal, Fishworks http://blogs.sun.com/ahl