Please enable JavaScript.
Coggle requires JavaScript to display documents.
cacheline tracking EEXIST - Coggle Diagram
cacheline tracking EEXIST
overview
same pages with shared cacheline appear in mutliple dma mapping
DMA_ATTR_SKIP_CPU_SYNC
how to investigate
add debug message
use bpftrace to re-write such check?
doesn't work
caused by lock?
possible reasons
such buffers are provided to IOs
shared READ IO buffers?
allocated kmalloc?
two buffers share same cache line
kmalloc buffer
missed dma unmap
not observed by bpftrace
same page added to bio twice
same page added to two bios?
two bios may be added in one req
report cases
raid1
RHEL-2817
rhel9.3
scsi error
comm:
mdX_raid1
observation
new entry is same with old stored entry in radix tree
the old entry is added before adding new one
old one is freed but not removed?
but remove check doesn't figure out the issue?
added otherwhere?
when old entry is added to tree?
how to
check in alloc, before clearing it
ahci
thin_check
direct io
/RHEL-2796
rhel9.3
nvme
BZ 2149635
blk_add_partitions
partition table read via page cache
'nvme' application
rhel8
rhel8