<html xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40"> <head> <meta http-equiv="Content-Type" content="text/html; charset=us-ascii"> <meta name="Generator" content="Microsoft Word 15 (filtered medium)"> <style></style> </head> <body lang="EN-IN" link="blue" vlink="purple" style="word-wrap:break-word"> <div class="WordSection1"> Hi Peter,<o:p></o:p> This code was just for test and reference. I primarily wanted to get review <o:p></o:p> on RFC. I have sent the cover letter explaining the issues as well as the <o:p></o:p> change. This patch is still in testing phase. Before moving forward I wanted <o:p></o:p> to confirm if my changes and the logic for port allocation looks good and get some history behind the earlier logic.<o:p></o:p> <o:p> </o:p> Attaching the doc explaining the issue and change just for reference.<o:p></o:p> PFA<o:p></o:p> <o:p> </o:p> Thanks<o:p></o:p> Regards<o:p></o:p> Divya Garg<o:p></o:p> <o:p> </o:p> <div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm"> From: Peter Krempa <pkrempa@redhat.com> Date: Wednesday, 24 November 2021 at 6:22 PM To: Divya Garg <divya.garg@nutanix.com> Cc: libvir-list@redhat.com <libvir-list@redhat.com>, Manish Mishra <manish.mishra@nutanix.com>, Prerna Saxena <prerna.saxena@nutanix.com> Subject: Re: [RFC 1/1] update index for serial device using taget.port<o:p></o:p> </div> <div> On Wed, Nov 24, 2021 at 07:06:06 -0500, divya wrote: Hi, just a few quick points, I'm not going to analyze what this does too deeply at this point. > From: root <root@localhost.localdomain> Author should be set properly. Also we require a proper commit message explaining the issue. Additionally we require that the submitter declares conformance with the developer certificate of origin: <a href="https://urldefense.proofpoint.com/v2/url?u=https-3A__www.libvirt.org_hacking.html-23developer-2Dcertificate-2Dof-2Dorigin&d=DwIBAg&c=s883GpUCOChKOHiocYtGcg&r=2QGHz-fTCVWImEBKe1ZcSe5t6UfasnhvdzD5DcixwOE&m=IUWwXppAjhBSqGVOlRTyZ70mhLrTUeeQA692o-Q942j1SPk8i8nrvQPV9KtRHOwl&s=23RVgvzwCkcam2r_OcJ3920vDnxkmIm4tY15Bb4LdFY&e=">https://urldefense.proofpoint.com/v2/url?u=https-3A__www.libvirt.org_hacking.html-23developer-2Dcertificate-2Dof-2Dorigin&d=DwIBAg&c=s883GpUCOChKOHiocYtGcg&r=2QGHz-fTCVWImEBKe1ZcSe5t6UfasnhvdzD5DcixwOE&m=IUWwXppAjhBSqGVOlRTyZ70mhLrTUeeQA692o-Q942j1SPk8i8nrvQPV9KtRHOwl&s=23RVgvzwCkcam2r_OcJ3920vDnxkmIm4tY15Bb4LdFY&e=</a> > > --- [...] > > diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c > index 5cfb2d91eb..edc5e897be 100644 > --- a/src/conf/domain_conf.c > +++ b/src/conf/domain_conf.c > @@ -65,6 +65,7 @@ > #include "virutil.h" > > #define VIR_FROM_THIS VIR_FROM_DOMAIN Firstly I see many problems with coding style: > +#define max_available_isa_serial_ports 4 We usually use uppercase macros > > VIR_LOG_INIT("conf.domain_conf"); > > @@ -19649,6 +19650,10 @@ virDomainDefParseXML(xmlXPathContextPtr ctxt, > g_autofree xmlNodePtr *nodes = NULL; > g_autofree char *tmp = NULL; > g_autoptr(virDomainDef) def = NULL; > + uint8_t used_serial_port_buffer = 0; We have virBitmap for such things. > + int isa_serial_count = 0; > + int next_available_serial_port = 0; > + int max_serial_port = -1; > > if (!(def = virDomainDefNew(xmlopt))) > return NULL; > @@ -19886,16 +19891,67 @@ virDomainDefParseXML(xmlXPathContextPtr ctxt, > if (!chr) > return NULL; > > - if (chr->target.port == -1) { > - int maxport = -1; > - for (j = 0; j < i; j++) { > - if (def->serials[j]->target.port > maxport) > - maxport = def->serials[j]->target.port; > + def->serials[def->nserials++] = chr; > + > + // Giving precedence to the isa-serial device since > + // only limited ports can be used for such devices. We don't use line comments (//) but block comments everywhere /* */ > + if (chr->targetType == VIR_DOMAIN_CHR_SERIAL_TARGET_MODEL_ISA_SERIAL) { > + // Taking the isa serial devices to start of the array. > + for (j = def->nserials; j > isa_serial_count; j--) > + def->serials[j] = def->serials[j-1]; > + def->serials[isa_serial_count++] = chr; > + } > + > + // Maintaining the buffer for first max_available_isa_serial_ports unused ports. > + if (chr->target.port != -1 && chr->target.port <= max_available_isa_serial_ports) { > + if (used_serial_port_buffer & (1 << chr->target.port)) { > + virReportError(VIR_ERR_INTERNAL_ERROR, > + _("target port [%d] already allocated."), > + chr->target.port); This is misaligned. > + return NULL; > } > - chr->target.port = maxport + 1; > + used_serial_port_buffer |= 1 << chr->target.port; > + } > + > + // Update max serial port used. > + if (chr->target.port > max_serial_port) > + max_serial_port = chr->target.port; > + } > + > + // Assign the ports to the devices. > + for (i = 0; i < n; i++) { > + if (def->serials[i]->target.port != -1) continue; > + // Assign one of the unused ports from first max_available_isa_serial_ports ports > + // to isa-serial device. > + if (def->serials[i]->targetType == VIR_DOMAIN_CHR_SERIAL_TARGET_MODEL_ISA_SERIAL) { > + > + // Search for the next available port. > + while (used_serial_port_buffer & (1 << next_available_serial_port) && > + next_available_serial_port <= max_available_isa_serial_ports) { > + next_available_serial_port++; > + } > + > + // qemu doesn't support more than max_available_isa_serial_ports isa devices. > + if (i > max_available_isa_serial_ports || > + next_available_serial_port > max_available_isa_serial_ports) { > + virReportError(VIR_ERR_INTERNAL_ERROR, > + _("Maximum supported number of ISA serial ports is %d."), > + max_available_isa_serial_ports); > + return NULL; > + } > + > + used_serial_port_buffer |= 1 << next_available_serial_port; > + def->serials[i]->target.port = next_available_serial_port; > + > + // Update max serial port used. > + if (def->serials[i]->target.port > max_serial_port) > + max_serial_port = def->serials[i]->target.port; > + > + } else { > + def->serials[i]->target.port = ++max_serial_port; > } > - def->serials[def->nserials++] = chr; > } In general, none of this code should be in the parser itself. Not even the existing code which you are changing actually. We have code which is meant to adjust and fill defaults for devices. For your case virDomainChrDefPostParse or virDomainDefPostParseCommon might be the right place. Note that until now the assignment code was rather trivial, but you are adding a massive amount of logic to this so it definitely doesn't belong to the parser any more. > + > VIR_FREE(nodes); > > if ((n = virXPathNodeSet("./devices/console", ctxt, &nodes)) < 0) { > diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c > index 7a185061d8..c8f8a27f30 100644 > --- a/src/qemu/qemu_command.c > +++ b/src/qemu/qemu_command.c > @@ -10947,11 +10947,21 @@ qemuBuildSerialChrDeviceProps(const virDomainDef *def, > return NULL; > } > > - if (virJSONValueObjectAdd(&props, > - "s:driver", virDomainChrSerialTargetModelTypeToString(serial->targetModel), > - "s:chardev", chardev, > - "s:id", serial->info.alias, > - NULL) < 0) > + if (serial->targetModel == VIR_DOMAIN_CHR_SERIAL_TARGET_MODEL_ISA_SERIAL && > + serial->target.port != -1) { > + if (virJSONValueObjectAdd(&props, > + "s:driver", virDomainChrSerialTargetModelTypeToString(serial->targetModel), > + "s:chardev", chardev, > + "s:id", serial->info.alias, > + "i:index", serial->target.port, You can use "k:index" which conditionally adds the 'index' field only when it's 0 or positive ... > + NULL) < 0) > + return NULL; > + } > + else if (virJSONValueObjectAdd(&props, > + "s:driver", virDomainChrSerialTargetModelTypeToString(serial->targetModel), > + "s:chardev", chardev, > + "s:id", serial->info.alias, > + NULL) < 0) ... so that you don't have to have this else branch here. Also all of this new code seems to be badly misalligned. > return NULL; > > if (qemuBuildDeviceAddressProps(props, def, &serial->info) < 0) > diff --git a/tests/qemuhotplugtest.c b/tests/qemuhotplugtest.c > index 263a92425c..88c9ea0339 100644 > --- a/tests/qemuhotplugtest.c > +++ b/tests/qemuhotplugtest.c > @@ -355,7 +355,6 @@ testQemuHotplug(const void *data) > if (keep) { > test->vm = vm; > } else { > - virObjectUnref(vm); > test->vm = NULL; > } > virDomainDeviceDefFree(dev); This is a very questionable hunk. Was this just for testing?<o:p></o:p> </div> </div> </body> </html>