diff options
author | Theo de Raadt <deraadt@cvs.openbsd.org> | 2005-12-29 09:20:05 +0000 |
---|---|---|
committer | Theo de Raadt <deraadt@cvs.openbsd.org> | 2005-12-29 09:20:05 +0000 |
commit | 27872770127f447c8bcdb355329991531408820d (patch) | |
tree | b50af9730cb1db3df8d5b48f6499020c18c57829 /libexec/ld.so/dir.c | |
parent | 56930fe9e7f02dba2acc6227f74ac1093d2d438f (diff) |
insanity debugged with djm -- here is the comment from the code:
+ * Some Maxim 1617 clones MAY NOT even read cmd 0xfc! When it is
+ * read, they will power-on-reset. Their default condition
+ * (control register bit 0x80) therefore will be that they assert
+ * /ALERT for the 5 potential errors that may occur. One of those
+ * errors is that the external temperature diode is missing. This
+ * is unfortunately a common choice of system designers, except
+ * suddenly now we get a /ALERT, which may on some chipsets cause
+ * us to receive an entirely unexpected SMI .. and then an NMI.
+ *
+ * As we probe each device, if we hit something which looks suspiciously
+ * like it may potentially be a 1617 or clone, we immediately set this
+ * variable to avoid reading that register offset.
Diffstat (limited to 'libexec/ld.so/dir.c')
0 files changed, 0 insertions, 0 deletions