On 10/29/20 7:53 AM, Mike Leach wrote:
Hi Suzuki,
On Wed, 28 Oct 2020 at 22:10, Suzuki K Poulose suzuki.poulose@arm.com wrote:
CoreSight ETMv4.4 obsoletes memory mapped access to ETM and mandates the system instructions for registers. This also implies that they may not be on the amba bus. Right now all the CoreSight components are accessed via memory map. Also, we have some common routines in coresight generic code driver (e.g, CS_LOCK, claim/disclaim), which assume the mmio. In order to preserve the generic algorithms at a single place and to allow dynamic switch for ETMs, this series introduces an abstraction layer for accessing a coresight device. It is designed such that the mmio access are fast tracked (i.e, without an indirect function call).
This will also help us to get rid of the driver+attribute specific sysfs show/store routines and replace them with a single routine to access a given register offset (which can be embedded in the dev_ext_attribute). This is not currently implemented in the series, but can be achieved.
Further we switch the generic routines to work with the abstraction. With this in place, we refactor the etm4x code a bit to allow for supporting the system instructions with very little new code. The changes also switch to using the system instructions by default even when we may have an MMIO.
We use TRCDEVARCH for the detection of the ETM component, which is a standard register as per CoreSight architecture, rather than the etm specific id register TRCIDR1. This is for making sure that we are able to detect the ETM via system instructions accurately, when the the trace unit could be anything (etm or a custom trace unit). To keep the backward compatibility for any existing broken impelementation which may not implement TRCDEVARCH, we fall back to TRCIDR1. Also this covers us for the changes in the future architecture [0].
The series has been mildly tested on a model for system instructions. I would really appreciate any testing on real hardware.
Applies on coresight/next.
[0] https://developer.arm.com/docs/ddi0601/g/aarch64-system-registers/trcidr1
Known issues: Checkpatch failure for "coresight: etm4x: Add sysreg access helpers" :
ERROR: Macros with complex values should be enclosed in parentheses #121: FILE: drivers/hwtracing/coresight/coresight-etm4x.h:153: +#define CASE_READ(res, x) \
case (x): { (res) = read_etm4x_sysreg_const_offset((x)); break; }
I don't know how to fix that without breaking the build ! Suggestions welcome.
I hit exactly the same issue with my recent v2 set. The checker really hates case statements in macros. In my case I shut it up for v3 by changing:-
#define MAPREG(cval, elem) \ case cval: \ reg->drv_store = &drvcfg->elem; \ break; if (((offset >= TRCEVENTCTL0R) && (offset <= TRCVIPCSSCTLR)) || ((offset >= TRCSEQRSTEVR) && (offset <= TRCEXTINSELR)) || ((offset >= TRCCIDCCTLR0) && (offset <= TRCVMIDCCTLR1))) { switch (offset) { /* 32 bit single control and filter registers */ MAPREG(TRCEVENTCTL0R, eventctrl0); MAPREG(TRCEVENTCTL1R, eventctrl1); MAPREG(TRCSTALLCTLR, stall_ctrl); MAPREG(TRCTSCTLR, ts_ctrl); MAPREG(TRCSYNCPR, syncfreq); MAPREG(TRCCCCTLR, ccctlr); MAPREG(TRCBBCTLR, bb_ctrl); MAPREG(TRCVICTLR, vinst_ctrl); MAPREG(TRCVIIECTLR, viiectlr); MAPREG(TRCVISSCTLR, vissctlr); MAPREG(TRCVIPCSSCTLR, vipcssctlr); MAPREG(TRCSEQRSTEVR, seq_rst); MAPREG(TRCSEQSTR, seq_state); MAPREG(TRCEXTINSELR, ext_inp); MAPREG(TRCCIDCCTLR0, ctxid_mask0); MAPREG(TRCCIDCCTLR1, ctxid_mask1); MAPREG(TRCVMIDCCTLR0, vmid_mask0); MAPREG(TRCVMIDCCTLR1, vmid_mask1); default: err = -EINVAL; break; }
to
err = -EINVAL;
#define CHECKREG(cval, elem) \ { \ if (offset == cval) { \ reg->drv_store = &drvcfg->elem; \ err = 0; \ break; \ } \ }
if (((offset >= TRCEVENTCTL0R) && (offset <= TRCVIPCSSCTLR)) || ((offset >= TRCSEQRSTEVR) && (offset <= TRCEXTINSELR)) || ((offset >= TRCCIDCCTLR0) && (offset <= TRCVMIDCCTLR1))) { do { CHECKREG(TRCEVENTCTL0R, eventctrl0); CHECKREG(TRCEVENTCTL1R, eventctrl1); CHECKREG(TRCSTALLCTLR, stall_ctrl); CHECKREG(TRCTSCTLR, ts_ctrl); CHECKREG(TRCSYNCPR, syncfreq); CHECKREG(TRCCCCTLR, ccctlr); CHECKREG(TRCBBCTLR, bb_ctrl); CHECKREG(TRCVICTLR, vinst_ctrl); CHECKREG(TRCVIIECTLR, viiectlr); CHECKREG(TRCVISSCTLR, vissctlr); CHECKREG(TRCVIPCSSCTLR, vipcssctlr); CHECKREG(TRCSEQRSTEVR, seq_rst); CHECKREG(TRCSEQSTR, seq_state); CHECKREG(TRCEXTINSELR, ext_inp); CHECKREG(TRCCIDCCTLR0, ctxid_mask0); CHECKREG(TRCCIDCCTLR1, ctxid_mask1); CHECKREG(TRCVMIDCCTLR0, vmid_mask0); CHECKREG(TRCVMIDCCTLR1, vmid_mask1); } while (0);
A bit contrived but functionally the same - and doesn't annoy the checker, No idea if the code is more or less efficient after compilation than a standard switch / case combo, but not on a critical path for me so wasn't bothered.
Thanks Mike. Functionally it is fine. But I think the Compiler wil be able to optimize it better with switch() for a large list of cases with consecutive values.
I guess there are more and more places that this is useful. I have Cc'ed the checkpatch maintainers in the patch in question.
Cheers Suzuki