From: Ira Weiny <weiny2-i2BcT+NCU+M@public.gmane.org>
To: "Hefty, Sean" <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Cc: Or Gerlitz <ogerlitz-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>,
Roland Dreier <roland-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>,
linux-rdma <linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>
Subject: Re: [PATCH 7/9] ib/pma: add include file for IBA performance counters definitions
Date: Wed, 15 Jun 2011 16:46:19 -0700 [thread overview]
Message-ID: <20110615164619.994a49a4.weiny2@llnl.gov> (raw)
In-Reply-To: <1828884A29C6694DAF28B7E6B8A8237302148F-P5GAC/sN6hmkrb+BlOpmy7fspsVTdybXVpNB7YpNyf8@public.gmane.org>
Minor issue.
PM is defined in the spec glossary to be "Partition Manager". :-( Therefore I have refrained from using "PM" or "pm" in structs or defines. I would not mind settling on an abbreviation; "PMA" perhaps?
Ira
On Wed, 15 Jun 2011 16:10:24 -0700
"Hefty, Sean" <sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org> wrote:
> Other than that one nit, the other patches in this series look good to me. See below for comments on this one.
>
> > +/*
> > + * PMA class portinfo capability mask bits
> > + */
> > +#define IB_PMA_CLASS_CAP_ALLPORTSELECT cpu_to_be16(1 << 8)
> > +#define IB_PMA_CLASS_CAP_EXT_WIDTH cpu_to_be16(1 << 9)
> > +#define IB_PMA_CLASS_CAP_XMIT_WAIT cpu_to_be16(1 << 12)
> > +
> > +#define IB_PMA_CLASS_PORT_INFO cpu_to_be16(0x0001)
>
> Hmm... we're defining attributes in different ways (see ib_sa.h and ib_smi.h). Someday we may want to make things consistent.
>
> > +#define IB_PMA_PORT_SAMPLES_CONTROL cpu_to_be16(0x0010)
> > +#define IB_PMA_PORT_SAMPLES_RESULT cpu_to_be16(0x0011)
> > +#define IB_PMA_PORT_COUNTERS cpu_to_be16(0x0012)
> > +#define IB_PMA_PORT_COUNTERS_EXT cpu_to_be16(0x001D)
> > +#define IB_PMA_PORT_SAMPLES_RESULT_EXT cpu_to_be16(0x001E)
> > +#define IB_PMA_PORT_COUNTERS_CONG cpu_to_be16(0xFF00)
> > +
> > +struct ib_perf {
> > + u8 base_version;
> > + u8 mgmt_class;
> > + u8 class_version;
> > + u8 method;
> > + __be16 status;
> > + __be16 unused;
> > + __be64 tid;
> > + __be16 attr_id;
> > + __be16 resv;
> > + __be32 attr_mod;
> > + u8 reserved[40];
> > + u8 data[192];
> > +} __attribute__ ((packed));
>
> This should use struct ib_mad_hdr, possibly adding appropriate values to ib_mad.h for the 40 and 192. See the enum in ib_mad.h with IB_MGMT_* values. I would also rename this to struct ib_pm_mad to better match the existing mad definitions.
>
> > +
> > +struct ib_pma_classportinfo {
> > + u8 base_version;
> > + u8 class_version;
> > + __be16 cap_mask;
> > + u8 reserved[3];
> > + u8 resp_time_value; /* only lower 5 bits */
> > + union ib_gid redirect_gid;
> > + __be32 redirect_tc_sl_fl; /* 8, 4, 20 bits respectively */
> > + __be16 redirect_lid;
> > + __be16 redirect_pkey;
> > + __be32 redirect_qp; /* only lower 24 bits */
> > + __be32 redirect_qkey;
> > + union ib_gid trap_gid;
> > + __be32 trap_tc_sl_fl; /* 8, 4, 20 bits respectively */
> > + __be16 trap_lid;
> > + __be16 trap_pkey;
> > + __be32 trap_hl_qp; /* 8, 24 bits respectively */
> > + __be32 trap_qkey;
> > +} __attribute__ ((packed));
>
> This needs to be replaced with struct ib_class_port_info in ib_mad.h.
>
> > +
> > +struct ib_pma_portsamplescontrol {
> > + u8 opcode;
> > + u8 port_select;
> > + u8 tick;
> > + u8 counter_width; /* only lower 3 bits */
> > + __be32 counter_mask0_9; /* 2, 10 * 3, bits */
> > + __be16 counter_mask10_14; /* 1, 5 * 3, bits */
>
> Comments for the bit fields aren't very clear, maybe say something more like '10 3-bit fields'. For things like counter_width, my preference is something like /* resv: 7:3, counter width: 2:0 */ to clarify which bits are used and indicate that the others are currently reserved.
>
> > + u8 sample_mechanisms;
> > + u8 sample_status; /* only lower 2 bits */
> > + __be64 option_mask;
> > + __be64 vendor_mask;
> > + __be32 sample_start;
> > + __be32 sample_interval;
> > + __be16 tag;
> > + __be16 counter_select[15];
> > +} __attribute__ ((packed));
>
> There are fields missing at the end: reserved 32-bits, SamplesOnlyOptionMask 64-bit, reserved 896-bits. This should make the structure naturally aligned, which could remove the packed attribute.
>
> > +struct ib_pma_portsamplesresult {
> > + __be16 tag;
> > + __be16 sample_status; /* only lower 2 bits */
> > + __be32 counter[15];
> > +} __attribute__ ((packed));
>
> this is naturally aligned
>
> > +struct ib_pma_portsamplesresult_ext {
> > + __be16 tag;
> > + __be16 sample_status; /* only lower 2 bits */
> > + __be32 extended_width; /* only upper 2 bits */
> > + __be64 counter[15];
> > +} __attribute__ ((packed));
>
> this is naturally aligned
>
> > +struct ib_pma_portcounters {
> > + u8 reserved;
> > + u8 port_select;
> > + __be16 counter_select;
> > + __be16 symbol_error_counter;
> > + u8 link_error_recovery_counter;
> > + u8 link_downed_counter;
> > + __be16 port_rcv_errors;
> > + __be16 port_rcv_remphys_errors;
> > + __be16 port_rcv_switch_relay_errors;
> > + __be16 port_xmit_discards;
> > + u8 port_xmit_constraint_errors;
> > + u8 port_rcv_constraint_errors;
> > + u8 reserved1;
> > + u8 lli_ebor_errors; /* 4, 4, bits */
>
> lli_ebor? what about link_overrun_errors instead, or at least something a little more descriptive?
>
> Enhancing the comment can also help: /* LocalLink: 7:4, BufferOverrun: 3:0 */
>
> > + __be16 reserved2;
> > + __be16 vl15_dropped;
> > + __be32 port_xmit_data;
> > + __be32 port_rcv_data;
> > + __be32 port_xmit_packets;
> > + __be32 port_rcv_packets;
> > +} __attribute__ ((packed));
>
> missing PortXmitWait 32-bits at end
>
> > +
> > +struct ib_pma_portcounters_cong {
> > + u8 reserved;
> > + u8 reserved1;
> > + __be16 port_check_rate;
> > + __be16 symbol_error_counter;
> > + u8 link_error_recovery_counter;
> > + u8 link_downed_counter;
> > + __be16 port_rcv_errors;
> > + __be16 port_rcv_remphys_errors;
> > + __be16 port_rcv_switch_relay_errors;
> > + __be16 port_xmit_discards;
> > + u8 port_xmit_constraint_errors;
> > + u8 port_rcv_constraint_errors;
> > + u8 reserved2;
> > + u8 lli_ebor_errors; /* 4, 4, bits */
> > + __be16 reserved3;
> > + __be16 vl15_dropped;
> > + __be64 port_xmit_data;
> > + __be64 port_rcv_data;
> > + __be64 port_xmit_packets;
> > + __be64 port_rcv_packets;
> > + __be64 port_xmit_wait;
> > + __be64 port_adr_events;
> > +} __attribute__ ((packed));
> > +
> > +#define IB_PMA_CONG_HW_CONTROL_TIMER 0x00
> > +#define IB_PMA_CONG_HW_CONTROL_SAMPLE 0x01
> >
> > +#define IB_PMA_SEL_SYMBOL_ERROR cpu_to_be16(0x0001)
> > +#define IB_PMA_SEL_LINK_ERROR_RECOVERY cpu_to_be16(0x0002)
> > +#define IB_PMA_SEL_LINK_DOWNED cpu_to_be16(0x0004)
> > +#define IB_PMA_SEL_PORT_RCV_ERRORS cpu_to_be16(0x0008)
> > +#define IB_PMA_SEL_PORT_RCV_REMPHYS_ERRORS cpu_to_be16(0x0010)
> > +#define IB_PMA_SEL_PORT_XMIT_DISCARDS cpu_to_be16(0x0040)
> > +#define IB_PMA_SEL_LOCAL_LINK_INTEGRITY_ERRORS cpu_to_be16(0x0200)
> > +#define IB_PMA_SEL_EXCESSIVE_BUFFER_OVERRUNS cpu_to_be16(0x0400)
> > +#define IB_PMA_SEL_PORT_VL15_DROPPED cpu_to_be16(0x0800)
> > +#define IB_PMA_SEL_PORT_XMIT_DATA cpu_to_be16(0x1000)
> > +#define IB_PMA_SEL_PORT_RCV_DATA cpu_to_be16(0x2000)
> > +#define IB_PMA_SEL_PORT_XMIT_PACKETS cpu_to_be16(0x4000)
> > +#define IB_PMA_SEL_PORT_RCV_PACKETS cpu_to_be16(0x8000)
> > +
> > +#define IB_PMA_SEL_CONG_ALL 0x01
> > +#define IB_PMA_SEL_CONG_PORT_DATA 0x02
> > +#define IB_PMA_SEL_CONG_XMIT 0x04
> > +#define IB_PMA_SEL_CONG_ROUTING 0x08
> > +
> > +struct ib_pma_portcounters_ext {
> > + u8 reserved;
> > + u8 port_select;
> > + __be16 counter_select;
> > + __be32 reserved1;
> > + __be64 port_xmit_data;
> > + __be64 port_rcv_data;
> > + __be64 port_xmit_packets;
> > + __be64 port_rcv_packets;
> > + __be64 port_unicast_xmit_packets;
> > + __be64 port_unicast_rcv_packets;
> > + __be64 port_multicast_xmit_packets;
> > + __be64 port_multicast_rcv_packets;
> > +} __attribute__ ((packed));
> > +
> > +#define IB_PMA_SELX_PORT_XMIT_DATA cpu_to_be16(0x0001)
> > +#define IB_PMA_SELX_PORT_RCV_DATA cpu_to_be16(0x0002)
> > +#define IB_PMA_SELX_PORT_XMIT_PACKETS cpu_to_be16(0x0004)
> > +#define IB_PMA_SELX_PORT_RCV_PACKETS cpu_to_be16(0x0008)
> > +#define IB_PMA_SELX_PORT_UNI_XMIT_PACKETS cpu_to_be16(0x0010)
> > +#define IB_PMA_SELX_PORT_UNI_RCV_PACKETS cpu_to_be16(0x0020)
> > +#define IB_PMA_SELX_PORT_MULTI_XMIT_PACKETS cpu_to_be16(0x0040)
> > +#define IB_PMA_SELX_PORT_MULTI_RCV_PACKETS cpu_to_be16(0x0080)
>
> Rather than having IB_PMA_SEL_* and IB_PMA_SELX_*, we could follow what's done in ib_sa.h:
>
> IB_PM_<structure>_<field>
>
> - Sean
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
Ira Weiny
Math Programmer/Computer Scientist
Lawrence Livermore National Lab
925-423-8008
weiny2-i2BcT+NCU+M@public.gmane.org
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2011-06-15 23:46 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-06-15 14:38 [PATCH 0/9] IBoE updates: GID change event and network flow counters Or Gerlitz
[not found] ` <alpine.LRH.2.00.1106151733410.15844-VYr5/9ddeaGSIdy2EShu12Xnswh1EIUO@public.gmane.org>
2011-06-15 14:39 ` [PATCH 1/9] ib/core: Add GID change event Or Gerlitz
2011-06-15 14:40 ` [PATCH 2/9] ib/mlx4: use GID change event on IBoE code Or Gerlitz
2011-06-15 14:41 ` [PATCH 3/9] mlx4: extend cap flags to 64 bit Or Gerlitz
2011-06-15 14:42 ` [PATCH 4/9] mlx4: align current extended capabilities to use the flags field Or Gerlitz
[not found] ` <alpine.LRH.2.00.1106151742330.15844-VYr5/9ddeaGSIdy2EShu12Xnswh1EIUO@public.gmane.org>
2011-07-07 19:19 ` [PATCH V1 " Or Gerlitz
2011-06-15 14:43 ` [PATCH 5/NNN] mlx4: Fix layout of QP context for correct location of counter index Or Gerlitz
2011-06-15 14:47 ` [PATCH 6/9] mlx4: add network flow counters Or Gerlitz
[not found] ` <alpine.LRH.2.00.1106151745060.15844-VYr5/9ddeaGSIdy2EShu12Xnswh1EIUO@public.gmane.org>
2011-06-15 15:22 ` Hefty, Sean
2011-06-15 14:48 ` [PATCH 7/9] ib/pma: add include file for IBA performance counters definitions Or Gerlitz
[not found] ` <alpine.LRH.2.00.1106151747390.15844-VYr5/9ddeaGSIdy2EShu12Xnswh1EIUO@public.gmane.org>
2011-06-15 16:08 ` Hal Rosenstock
[not found] ` <4DF8D90A.1000208-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2011-06-16 8:01 ` Or Gerlitz
2011-06-15 23:10 ` Hefty, Sean
[not found] ` <1828884A29C6694DAF28B7E6B8A8237302148F-P5GAC/sN6hmkrb+BlOpmy7fspsVTdybXVpNB7YpNyf8@public.gmane.org>
2011-06-15 23:46 ` Ira Weiny [this message]
[not found] ` <20110615164619.994a49a4.weiny2-i2BcT+NCU+M@public.gmane.org>
2011-06-15 23:51 ` Hefty, Sean
2011-06-16 7:39 ` Or Gerlitz
[not found] ` <4DF9B319.6070100-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2011-06-16 16:19 ` Hefty, Sean
[not found] ` <1828884A29C6694DAF28B7E6B8A8237302184A-P5GAC/sN6hmkrb+BlOpmy7fspsVTdybXVpNB7YpNyf8@public.gmane.org>
2011-06-16 17:57 ` Or Gerlitz
2011-06-20 16:46 ` IBoE updates - GID change event / network flow counters (was: [PATCH 7/9] ib/pma: add include file for IBA perf counters) Or Gerlitz
[not found] ` <4DFF797E.5090904-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2011-06-20 18:18 ` Roland Dreier
[not found] ` <BANLkTi=VR6v=xSwzLRHDL1x2=05wF-H5KQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2011-06-20 20:38 ` Or Gerlitz
2011-07-16 19:57 ` Or Gerlitz
2011-06-16 8:20 ` [PATCH V1 7/9] ib/pma: add include file for IBA performance counters definitions Or Gerlitz
[not found] ` <alpine.LRH.2.00.1106161119120.12552-VYr5/9ddeaGSIdy2EShu12Xnswh1EIUO@public.gmane.org>
2011-06-16 14:18 ` Or Gerlitz
[not found] ` <4DFA10CC.5080002-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2011-06-16 15:49 ` Mike Marciniszyn
[not found] ` <35AAF1E4A771E142979F27B51793A4888838E71F03-HolNjIBXvBOXx9kJd3VG2h2eb7JE58TQ@public.gmane.org>
2011-06-16 18:14 ` Or Gerlitz
[not found] ` <BANLkTikpMWYaXB8Ne+FaLGHi2L5TUCNDkQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2011-07-05 15:49 ` Mike Marciniszyn
[not found] ` <35AAF1E4A771E142979F27B51793A4888838FEE740-HolNjIBXvBOXx9kJd3VG2h2eb7JE58TQ@public.gmane.org>
2011-07-05 16:01 ` Or Gerlitz
[not found] ` <4E133575.8040105-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2011-07-05 19:34 ` Roland Dreier
[not found] ` <CAL1RGDX7U=PMjWadeS3Gp+xST56xJiK=qydwu1PMcvkfb7DcnA-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2011-07-06 3:11 ` Or Gerlitz
[not found] ` <CAJZOPZK3tT9gXP46ek-tO3xyvj7=eKyGJXBoim9uOMbJc4BtfQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2011-07-06 12:39 ` Mike Marciniszyn
[not found] ` <35AAF1E4A771E142979F27B51793A4888838FEE894-HolNjIBXvBOXx9kJd3VG2h2eb7JE58TQ@public.gmane.org>
2011-07-06 15:51 ` Roland Dreier
[not found] ` <CAL1RGDWHYMQHbZWwkysqZ7TLrB_ZDJWSTC4YCqk1wpme_3+BrQ-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2011-07-06 16:57 ` Hal Rosenstock
[not found] ` <4E14940A.9080700-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2011-07-06 17:50 ` Roland Dreier
[not found] ` <CAL1RGDVpuwsgfVRxNw098W_wRECGn0uinQR+8txE96n5KtDXgw-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2011-07-06 18:15 ` Jason Gunthorpe
[not found] ` <20110706181507.GC1162-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2011-07-06 18:24 ` Hal Rosenstock
[not found] ` <4E14A87B.2070405-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2011-07-06 18:45 ` Jason Gunthorpe
[not found] ` <20110706184511.GD1162-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2011-07-06 18:53 ` Hal Rosenstock
2011-07-06 18:27 ` Hal Rosenstock
2011-07-07 6:55 ` Or Gerlitz
2011-07-06 16:47 ` Ira Weiny
2011-07-05 15:46 ` [PATCH V2 " Or Gerlitz
2011-06-15 14:49 ` [PATCH 8/9] ib/mlx4: use flow counters on IBoE ports Or Gerlitz
2011-06-15 14:51 ` [PATCH 9/9] ib/mlx4: support PMA counters for IBoE Or Gerlitz
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20110615164619.994a49a4.weiny2@llnl.gov \
--to=weiny2-i2bct+ncu+m@public.gmane.org \
--cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=ogerlitz-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org \
--cc=roland-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org \
--cc=sean.hefty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox